shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity

[Clang][SVE] Change LLVM representation of ACLE tuple types to be struct based. (#108008)

Browse Source 
This implements our original design now that LLVM is comfortable with
structs and arrays of scalable vector types. All SVE ACLE intrinsics
already use struct types so the effect of this change is purely the
types used for alloca and function parameters.
    
There should be no C/C++ user visible change with this patch.
This commit is contained in:
Paul Walker 2024-09-13 12:45:21 +01:00 committed by GitHub
parent 7574e1ddc4
commit 992a64aad3
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
108 changed files with 27789 additions and 22511 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
clang/include/clang/Basic/AArch64SVEACLETypes.def
View File

@ -46,9 +46,8 @@
//
// - ElBits is the size of one element in bits.
//
// - NF enumerates the number of sub-vectors.
// TODO: Tuple types are represented as a concatination of "NumEls x ElBits"
// vectors. This will be changed to become a struct containing NF vectors.
// - NF enumerates the number of vectors whereby 1 implies a single vector,
// with other values implying a struct of NF "NumEls x NumEls" vectors.
//
// - IsSigned is true for vectors of signed integer elements and
// for vectors of floating-point elements.

109
clang/lib/CodeGen/CGBuiltin.cpp
View File

@ -9990,31 +9990,6 @@ Value *CodeGenFunction::EmitSVEStructLoad(const SVETypeFlags &TypeFlags,
SmallVectorImpl<Value*> &Ops,
unsigned IntID) {
llvm::ScalableVectorType *VTy = getSVEType(TypeFlags);
unsigned N;
switch (IntID) {
case Intrinsic::aarch64_sve_ld2_sret:
case Intrinsic::aarch64_sve_ld1_pn_x2:
case Intrinsic::aarch64_sve_ldnt1_pn_x2:
case Intrinsic::aarch64_sve_ld2q_sret:
N = 2;
break;
case Intrinsic::aarch64_sve_ld3_sret:
case Intrinsic::aarch64_sve_ld3q_sret:
N = 3;
break;
case Intrinsic::aarch64_sve_ld4_sret:
case Intrinsic::aarch64_sve_ld1_pn_x4:
case Intrinsic::aarch64_sve_ldnt1_pn_x4:
case Intrinsic::aarch64_sve_ld4q_sret:
N = 4;
break;
default:
llvm_unreachable("unknown intrinsic!");
}
auto RetTy = llvm::VectorType::get(VTy->getElementType(),
VTy->getElementCount() * N);
Value *Predicate = EmitSVEPredicateCast(Ops[0], VTy);
Value *BasePtr = Ops[1];
@ -10023,15 +9998,7 @@ Value *CodeGenFunction::EmitSVEStructLoad(const SVETypeFlags &TypeFlags,
BasePtr = Builder.CreateGEP(VTy, BasePtr, Ops[2]);
Function *F = CGM.getIntrinsic(IntID, {VTy});
Value *Call = Builder.CreateCall(F, {Predicate, BasePtr});
unsigned MinElts = VTy->getMinNumElements();
Value *Ret = llvm::PoisonValue::get(RetTy);
for (unsigned I = 0; I < N; I++) {
Value *Idx = ConstantInt::get(CGM.Int64Ty, I * MinElts);
Value *SRet = Builder.CreateExtractValue(Call, I);
Ret = Builder.CreateInsertVector(RetTy, Ret, SRet, Idx);
}
return Ret;
return Builder.CreateCall(F, {Predicate, BasePtr});
}
Value *CodeGenFunction::EmitSVEStructStore(const SVETypeFlags &TypeFlags,
@ -10304,6 +10271,19 @@ Value *CodeGenFunction::EmitSVEReinterpret(Value *Val, llvm::Type *Ty) {
// view (when storing/reloading), whereas the svreinterpret builtin
// implements bitwise equivalent cast from register point of view.
// LLVM CodeGen for a bitcast must add an explicit REV for big-endian.
if (auto *StructTy = dyn_cast<StructType>(Ty)) {
Value *Tuple = llvm::PoisonValue::get(Ty);
for (unsigned I = 0; I < StructTy->getNumElements(); ++I) {
Value *In = Builder.CreateExtractValue(Val, I);
Value *Out = Builder.CreateBitCast(In, StructTy->getTypeAtIndex(I));
Tuple = Builder.CreateInsertValue(Tuple, Out, I);
}
return Tuple;
}
return Builder.CreateBitCast(Val, Ty);
}
@ -10346,44 +10326,26 @@ CodeGenFunction::getSVEOverloadTypes(const SVETypeFlags &TypeFlags,
}
Value *CodeGenFunction::EmitSVETupleSetOrGet(const SVETypeFlags &TypeFlags,
llvm::Type *Ty,
ArrayRef<Value *> Ops) {
assert((TypeFlags.isTupleSet() || TypeFlags.isTupleGet()) &&
"Expects TypleFlags.isTupleSet() or TypeFlags.isTupleGet()");
unsigned I = cast<ConstantInt>(Ops[1])->getSExtValue();
auto *SingleVecTy = dyn_cast<llvm::ScalableVectorType>(
TypeFlags.isTupleSet() ? Ops[2]->getType() : Ty);
if (!SingleVecTy)
return nullptr;
Value *Idx = ConstantInt::get(CGM.Int64Ty,
I * SingleVecTy->getMinNumElements());
unsigned Idx = cast<ConstantInt>(Ops[1])->getZExtValue();
if (TypeFlags.isTupleSet())
return Builder.CreateInsertVector(Ty, Ops[0], Ops[2], Idx);
return Builder.CreateExtractVector(Ty, Ops[0], Idx);
return Builder.CreateInsertValue(Ops[0], Ops[2], Idx);
return Builder.CreateExtractValue(Ops[0], Idx);
}
Value *CodeGenFunction::EmitSVETupleCreate(const SVETypeFlags &TypeFlags,
llvm::Type *Ty,
ArrayRef<Value *> Ops) {
llvm::Type *Ty,
ArrayRef<Value *> Ops) {
assert(TypeFlags.isTupleCreate() && "Expects TypleFlag isTupleCreate");
auto *SrcTy = dyn_cast<llvm::ScalableVectorType>(Ops[0]->getType());
Value *Tuple = llvm::PoisonValue::get(Ty);
for (unsigned Idx = 0; Idx < Ops.size(); Idx++)
Tuple = Builder.CreateInsertValue(Tuple, Ops[Idx], Idx);
if (!SrcTy)
return nullptr;
unsigned MinElts = SrcTy->getMinNumElements();
Value *Call = llvm::PoisonValue::get(Ty);
for (unsigned I = 0; I < Ops.size(); I++) {
Value *Idx = ConstantInt::get(CGM.Int64Ty, I * MinElts);
Call = Builder.CreateInsertVector(Ty, Call, Ops[I], Idx);
}
return Call;
return Tuple;
}
Value *CodeGenFunction::FormSVEBuiltinResult(Value *Call) {
@ -10453,27 +10415,14 @@ void CodeGenFunction::GetAArch64SVEProcessedOperands(
continue;
}
if (IsTupleGetOrSet || !isa<ScalableVectorType>(Arg->getType())) {
Ops.push_back(Arg);
if (isa<StructType>(Arg->getType()) && !IsTupleGetOrSet) {
for (unsigned I = 0; I < Arg->getType()->getStructNumElements(); ++I)
Ops.push_back(Builder.CreateExtractValue(Arg, I));
continue;
}
auto *VTy = cast<ScalableVectorType>(Arg->getType());
unsigned MinElts = VTy->getMinNumElements();
bool IsPred = VTy->getElementType()->isIntegerTy(1);
unsigned N = (MinElts * VTy->getScalarSizeInBits()) / (IsPred ? 16 : 128);
if (N == 1) {
Ops.push_back(Arg);
continue;
}
for (unsigned I = 0; I < N; ++I) {
Value *Idx = ConstantInt::get(CGM.Int64Ty, (I * MinElts) / N);
auto *NewVTy =
ScalableVectorType::get(VTy->getElementType(), MinElts / N);
Ops.push_back(Builder.CreateExtractVector(NewVTy, Arg, Idx));
}
Ops.push_back(Arg);
}
}
@ -10511,7 +10460,7 @@ Value *CodeGenFunction::EmitAArch64SVEBuiltinExpr(unsigned BuiltinID,
else if (TypeFlags.isStructStore())
return EmitSVEStructStore(TypeFlags, Ops, Builtin->LLVMIntrinsic);
else if (TypeFlags.isTupleSet() || TypeFlags.isTupleGet())
return EmitSVETupleSetOrGet(TypeFlags, Ty, Ops);
return EmitSVETupleSetOrGet(TypeFlags, Ops);
else if (TypeFlags.isTupleCreate())
return EmitSVETupleCreate(TypeFlags, Ty, Ops);
else if (TypeFlags.isUndef())

1
clang/lib/CodeGen/CodeGenFunction.h
View File

@ -4628,7 +4628,6 @@ public:
llvm::ScalableVectorType *getSVEType(const SVETypeFlags &TypeFlags);
llvm::ScalableVectorType *getSVEPredType(const SVETypeFlags &TypeFlags);
llvm::Value *EmitSVETupleSetOrGet(const SVETypeFlags &TypeFlags,
llvm::Type *ReturnType,
ArrayRef<llvm::Value *> Ops);
llvm::Value *EmitSVETupleCreate(const SVETypeFlags &TypeFlags,
llvm::Type *ReturnType,

17
clang/lib/CodeGen/CodeGenTypes.cpp
View File

@ -509,9 +509,20 @@ llvm::Type *CodeGenTypes::ConvertType(QualType T) {
{
ASTContext::BuiltinVectorTypeInfo Info =
Context.getBuiltinVectorTypeInfo(cast<BuiltinType>(Ty));
return llvm::ScalableVectorType::get(ConvertType(Info.ElementType),
Info.EC.getKnownMinValue() *
Info.NumVectors);
auto VTy =
llvm::VectorType::get(ConvertType(Info.ElementType), Info.EC);
switch (Info.NumVectors) {
default:
llvm_unreachable("Expected 1, 2, 3 or 4 vectors!");
case 1:
return VTy;
case 2:
return llvm::StructType::get(VTy, VTy);
case 3:
return llvm::StructType::get(VTy, VTy, VTy);
case 4:
return llvm::StructType::get(VTy, VTy, VTy, VTy);
}
}
case BuiltinType::SveCount:
return llvm::TargetExtType::get(getLLVMContext(), "aarch64.svcount");

328
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_add.c
View File

@ -24,16 +24,12 @@
// x2
// CHECK-LABEL: @test_svadd_write_single2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svadd_write_single2_s32j11svint32x2_tu11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_single2_s32(uint32_t slice_base, svint32x2_t zn, svint32_t zm) __arm_streaming __arm_inout("za") {
@ -42,16 +38,12 @@ void test_svadd_write_single2_s32(uint32_t slice_base, svint32x2_t zn, svint32_t
// CHECK-LABEL: @test_svadd_write_single2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svadd_write_single2_u32j12svuint32x2_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_single2_u32(uint32_t slice_base, svuint32x2_t zn, svuint32_t zm) __arm_streaming __arm_inout("za") {
@ -60,16 +52,12 @@ void test_svadd_write_single2_u32(uint32_t slice_base, svuint32x2_t zn, svuint32
// CHECK-LABEL: @test_svadd_write_single2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svadd_write_single2_s64j11svint64x2_tu11__SVInt64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_single2_s64(uint32_t slice_base, svint64x2_t zn, svint64_t zm) __arm_streaming __arm_inout("za") {
@ -78,16 +66,12 @@ void test_svadd_write_single2_s64(uint32_t slice_base, svint64x2_t zn, svint64_t
// CHECK-LABEL: @test_svadd_write_single2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svadd_write_single2_u64j12svuint64x2_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_single2_u64(uint32_t slice_base, svuint64x2_t zn, svuint64_t zm) __arm_streaming __arm_inout("za") {
@ -98,20 +82,12 @@ void test_svadd_write_single2_u64(uint32_t slice_base, svuint64x2_t zn, svuint64
// CHECK-LABEL: @test_svadd_write_single4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svadd_write_single4_s32j11svint32x4_tu11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_single4_s32(uint32_t slice_base, svint32x4_t zn, svint32_t zm) __arm_streaming __arm_inout("za") {
@ -120,20 +96,12 @@ void test_svadd_write_single4_s32(uint32_t slice_base, svint32x4_t zn, svint32_t
// CHECK-LABEL: @test_svadd_write_single4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svadd_write_single4_u32j12svuint32x4_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_single4_u32(uint32_t slice_base, svuint32x4_t zn, svuint32_t zm) __arm_streaming __arm_inout("za") {
@ -142,20 +110,12 @@ void test_svadd_write_single4_u32(uint32_t slice_base, svuint32x4_t zn, svuint32
// CHECK-LABEL: @test_svadd_write_single4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svadd_write_single4_s64j11svint64x4_tu11__SVInt64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_single4_s64(uint32_t slice_base, svint64x4_t zn, svint64_t zm) __arm_streaming __arm_inout("za") {
@ -164,20 +124,12 @@ void test_svadd_write_single4_s64(uint32_t slice_base, svint64x4_t zn, svint64_t
// CHECK-LABEL: @test_svadd_write_single4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svadd_write_single4_u64j12svuint64x4_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_single4_u64(uint32_t slice_base, svuint64x4_t zn, svuint64_t zm) __arm_streaming __arm_inout("za") {
@ -192,20 +144,12 @@ void test_svadd_write_single4_u64(uint32_t slice_base, svuint64x4_t zn, svuint64
// CHECK-LABEL: @test_svadd_write_multi2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svadd_write_multi2_s32j11svint32x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_multi2_s32(uint32_t slice_base, svint32x2_t zn, svint32x2_t zm) __arm_streaming __arm_inout("za") {
@ -214,20 +158,12 @@ void test_svadd_write_multi2_s32(uint32_t slice_base, svint32x2_t zn, svint32x2_
// CHECK-LABEL: @test_svadd_write_multi2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svadd_write_multi2_u32j12svuint32x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_multi2_u32(uint32_t slice_base, svuint32x2_t zn, svuint32x2_t zm) __arm_streaming __arm_inout("za") {
@ -236,20 +172,12 @@ void test_svadd_write_multi2_u32(uint32_t slice_base, svuint32x2_t zn, svuint32x
// CHECK-LABEL: @test_svadd_write_multi2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svadd_write_multi2_s64j11svint64x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_multi2_s64(uint32_t slice_base, svint64x2_t zn, svint64x2_t zm) __arm_streaming __arm_inout("za") {
@ -258,20 +186,12 @@ void test_svadd_write_multi2_s64(uint32_t slice_base, svint64x2_t zn, svint64x2_
// CHECK-LABEL: @test_svadd_write_multi2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svadd_write_multi2_u64j12svuint64x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_multi2_u64(uint32_t slice_base, svuint64x2_t zn, svuint64x2_t zm) __arm_streaming __arm_inout("za") {
@ -282,28 +202,12 @@ void test_svadd_write_multi2_u64(uint32_t slice_base, svuint64x2_t zn, svuint64x
// CHECK-LABEL: @test_svadd_write_multi4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE2:%.*]], <vscale x 4 x i32> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svadd_write_multi4_s32j11svint32x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE2:%.*]], <vscale x 4 x i32> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_multi4_s32(uint32_t slice_base, svint32x4_t zn, svint32x4_t zm) __arm_streaming __arm_inout("za") {
@ -312,28 +216,12 @@ void test_svadd_write_multi4_s32(uint32_t slice_base, svint32x4_t zn, svint32x4_
// CHECK-LABEL: @test_svadd_write_multi4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE2:%.*]], <vscale x 4 x i32> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svadd_write_multi4_u32j12svuint32x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE2:%.*]], <vscale x 4 x i32> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_multi4_u32(uint32_t slice_base, svuint32x4_t zn, svuint32x4_t zm) __arm_streaming __arm_inout("za") {
@ -342,28 +230,12 @@ void test_svadd_write_multi4_u32(uint32_t slice_base, svuint32x4_t zn, svuint32x
// CHECK-LABEL: @test_svadd_write_multi4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE2:%.*]], <vscale x 2 x i64> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svadd_write_multi4_s64j11svint64x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE2:%.*]], <vscale x 2 x i64> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_multi4_s64(uint32_t slice_base, svint64x4_t zn, svint64x4_t zm) __arm_streaming __arm_inout("za") {
@ -372,28 +244,12 @@ void test_svadd_write_multi4_s64(uint32_t slice_base, svint64x4_t zn, svint64x4_
// CHECK-LABEL: @test_svadd_write_multi4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE2:%.*]], <vscale x 2 x i64> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svadd_write_multi4_u64j12svuint64x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE2:%.*]], <vscale x 2 x i64> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_write_multi4_u64(uint32_t slice_base, svuint64x4_t zn, svuint64x4_t zm) __arm_streaming __arm_inout("za") {
@ -408,16 +264,12 @@ void test_svadd_write_multi4_u64(uint32_t slice_base, svuint64x4_t zn, svuint64x
// CHECK-LABEL: @test_svadd_za32_vg1x2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za32_vg1x2_f32j13svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za32_vg1x2_f32(uint32_t slice_base, svfloat32x2_t zn) __arm_streaming __arm_inout("za") {
@ -426,16 +278,12 @@ void test_svadd_za32_vg1x2_f32(uint32_t slice_base, svfloat32x2_t zn) __arm_stre
// CHECK-LABEL: @test_svadd_za32_vg1x2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za32_vg1x2_s32j11svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za32_vg1x2_s32(uint32_t slice_base, svint32x2_t zn) __arm_streaming __arm_inout("za") {
@ -444,16 +292,12 @@ void test_svadd_za32_vg1x2_s32(uint32_t slice_base, svint32x2_t zn) __arm_stream
// CHECK-LABEL: @test_svadd_za32_vg1x2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za32_vg1x2_u32j12svuint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za32_vg1x2_u32(uint32_t slice_base, svuint32x2_t zn) __arm_streaming __arm_inout("za") {
@ -462,16 +306,12 @@ void test_svadd_za32_vg1x2_u32(uint32_t slice_base, svuint32x2_t zn) __arm_strea
// CHECK-LABEL: @test_svadd_za64_vg1x2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za64_vg1x2_f64j13svfloat64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za64_vg1x2_f64(uint32_t slice_base, svfloat64x2_t zn) __arm_streaming __arm_inout("za") {
@ -480,16 +320,12 @@ void test_svadd_za64_vg1x2_f64(uint32_t slice_base, svfloat64x2_t zn) __arm_stre
// CHECK-LABEL: @test_svadd_za64_vg1x2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za64_vg1x2_s64j11svint64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za64_vg1x2_s64(uint32_t slice_base, svint64x2_t zn) __arm_streaming __arm_inout("za") {
@ -498,16 +334,12 @@ void test_svadd_za64_vg1x2_s64(uint32_t slice_base, svint64x2_t zn) __arm_stream
// CHECK-LABEL: @test_svadd_za64_vg1x2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za64_vg1x2_u64j12svuint64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za64_vg1x2_u64(uint32_t slice_base, svuint64x2_t zn) __arm_streaming __arm_inout("za") {
@ -518,20 +350,12 @@ void test_svadd_za64_vg1x2_u64(uint32_t slice_base, svuint64x2_t zn) __arm_strea
// CHECK-LABEL: @test_svadd_za32_vg1x4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za32_vg1x4_f32j13svfloat32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za32_vg1x4_f32(uint32_t slice_base, svfloat32x4_t zn) __arm_streaming __arm_inout("za") {
@ -540,20 +364,12 @@ void test_svadd_za32_vg1x4_f32(uint32_t slice_base, svfloat32x4_t zn) __arm_stre
// CHECK-LABEL: @test_svadd_za32_vg1x4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za32_vg1x4_s32j11svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za32_vg1x4_s32(uint32_t slice_base, svint32x4_t zn) __arm_streaming __arm_inout("za") {
@ -562,20 +378,12 @@ void test_svadd_za32_vg1x4_s32(uint32_t slice_base, svint32x4_t zn) __arm_stream
// CHECK-LABEL: @test_svadd_za32_vg1x4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za32_vg1x4_u32j12svuint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za32_vg1x4_u32(uint32_t slice_base, svuint32x4_t zn) __arm_streaming __arm_inout("za") {
@ -584,20 +392,12 @@ void test_svadd_za32_vg1x4_u32(uint32_t slice_base, svuint32x4_t zn) __arm_strea
// CHECK-LABEL: @test_svadd_za64_vg1x4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za64_vg1x4_f64j13svfloat64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za64_vg1x4_f64(uint32_t slice_base, svfloat64x4_t zn) __arm_streaming __arm_inout("za") {
@ -606,20 +406,12 @@ void test_svadd_za64_vg1x4_f64(uint32_t slice_base, svfloat64x4_t zn) __arm_stre
// CHECK-LABEL: @test_svadd_za64_vg1x4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za64_vg1x4_s64j11svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za64_vg1x4_s64(uint32_t slice_base, svint64x4_t zn) __arm_streaming __arm_inout("za") {
@ -628,20 +420,12 @@ void test_svadd_za64_vg1x4_s64(uint32_t slice_base, svint64x4_t zn) __arm_stream
// CHECK-LABEL: @test_svadd_za64_vg1x4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svadd_za64_vg1x4_u64j12svuint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svadd_za64_vg1x4_u64(uint32_t slice_base, svuint64x4_t zn) __arm_streaming __arm_inout("za") {

112
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_add_sub_za16.c
View File

@ -17,19 +17,15 @@
#endif
// CHECK-LABEL: define dso_local void @test_svadd_za16_vg1x2_f16(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z25test_svadd_za16_vg1x2_f16j13svfloat16x2_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]])
// CHECK-CXX-NEXT: ret void
//
void test_svadd_za16_vg1x2_f16(uint32_t slice, svfloat16x2_t zn) __arm_streaming __arm_inout("za") {
@ -37,23 +33,15 @@ void test_svadd_za16_vg1x2_f16(uint32_t slice, svfloat16x2_t zn) __arm_streaming
}
// CHECK-LABEL: define dso_local void @test_svadd_za16_vg1x4_f16(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z25test_svadd_za16_vg1x4_f16j13svfloat16x4_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]])
// CHECK-CXX-NEXT: ret void
//
void test_svadd_za16_vg1x4_f16(uint32_t slice, svfloat16x4_t zn) __arm_streaming __arm_inout("za") {
@ -61,19 +49,15 @@ void test_svadd_za16_vg1x4_f16(uint32_t slice, svfloat16x4_t zn) __arm_streaming
}
// CHECK-LABEL: define dso_local void @test_svsub_za16_vg1x2_f16(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z25test_svsub_za16_vg1x2_f16j13svfloat16x2_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]])
// CHECK-CXX-NEXT: ret void
//
void test_svsub_za16_vg1x2_f16(uint32_t slice, svfloat16x2_t zn) __arm_streaming __arm_inout("za") {
@ -81,23 +65,15 @@ void test_svsub_za16_vg1x2_f16(uint32_t slice, svfloat16x2_t zn) __arm_streaming
}
// CHECK-LABEL: define dso_local void @test_svsub_za16_vg1x4_f16(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z25test_svsub_za16_vg1x4_f16j13svfloat16x4_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]])
// CHECK-CXX-NEXT: ret void
//
void test_svsub_za16_vg1x4_f16(uint32_t slice, svfloat16x4_t zn) __arm_streaming __arm_inout("za") {
@ -105,19 +81,15 @@ void test_svsub_za16_vg1x4_f16(uint32_t slice, svfloat16x4_t zn) __arm_streaming
}
// CHECK-LABEL: define dso_local void @test_svadd_za16_vg1x2_bf16(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z26test_svadd_za16_vg1x2_bf16j14svbfloat16x2_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]])
// CHECK-CXX-NEXT: ret void
//
void test_svadd_za16_vg1x2_bf16(uint32_t slice, svbfloat16x2_t zn) __arm_streaming __arm_inout("za") {
@ -125,23 +97,15 @@ void test_svadd_za16_vg1x2_bf16(uint32_t slice, svbfloat16x2_t zn) __arm_streami
}
// CHECK-LABEL: define dso_local void @test_svadd_za16_vg1x4_bf16(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z26test_svadd_za16_vg1x4_bf16j14svbfloat16x4_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]])
// CHECK-CXX-NEXT: ret void
//
void test_svadd_za16_vg1x4_bf16(uint32_t slice, svbfloat16x4_t zn) __arm_streaming __arm_inout("za") {
@ -149,19 +113,15 @@ void test_svadd_za16_vg1x4_bf16(uint32_t slice, svbfloat16x4_t zn) __arm_streami
}
// CHECK-LABEL: define dso_local void @test_svsub_za16_vg1x2_bf16(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z26test_svsub_za16_vg1x2_bf16j14svbfloat16x2_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]])
// CHECK-CXX-NEXT: ret void
//
void test_svsub_za16_vg1x2_bf16(uint32_t slice, svbfloat16x2_t zn) __arm_streaming __arm_inout("za") {
@ -169,23 +129,15 @@ void test_svsub_za16_vg1x2_bf16(uint32_t slice, svbfloat16x2_t zn) __arm_streami
}
// CHECK-LABEL: define dso_local void @test_svsub_za16_vg1x4_bf16(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z26test_svsub_za16_vg1x4_bf16j14svbfloat16x4_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]])
// CHECK-CXX-NEXT: ret void
//
void test_svsub_za16_vg1x4_bf16(uint32_t slice, svbfloat16x4_t zn) __arm_streaming __arm_inout("za") {

1056
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_clamp.c
View File

File diff suppressed because it is too large Load Diff

518
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_cvt.c
View File

@ -19,17 +19,13 @@
// CHECK-LABEL: @test_cvt_f16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z15test_cvt_f16_x213svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
//
svfloat16_t test_cvt_f16_x2(svfloat32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvt_f16,_f32_x2,,)(zn);
@ -37,17 +33,13 @@ svfloat16_t test_cvt_f16_x2(svfloat32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_cvt_bf16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.bfcvt.x2(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.bfcvt.x2(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z16test_cvt_bf16_x213svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.bfcvt.x2(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.bfcvt.x2(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
//
svbfloat16_t test_cvt_bf16_x2(svfloat32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvt_bf16,_f32_x2,,)(zn);
@ -57,25 +49,27 @@ svbfloat16_t test_cvt_bf16_x2(svfloat32x2_t zn) __arm_streaming {
// x2
// CHECK-LABEL: @test_svcvt_f32_u32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ucvtf.x2.nxv4f32.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ucvtf.x2.nxv4f32.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z21test_svcvt_f32_u32_x212svuint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ucvtf.x2.nxv4f32.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ucvtf.x2.nxv4f32.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svcvt_f32_u32_x2(svuint32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvt_f32,_u32_x2,,)(zn);
@ -83,25 +77,27 @@ svfloat32x2_t test_svcvt_f32_u32_x2(svuint32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svcvt_f32_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.scvtf.x2.nxv4f32.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.scvtf.x2.nxv4f32.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z21test_svcvt_f32_s32_x211svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.scvtf.x2.nxv4f32.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.scvtf.x2.nxv4f32.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svcvt_f32_s32_x2(svint32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvt_f32,_s32_x2,,)(zn);
@ -109,25 +105,27 @@ svfloat32x2_t test_svcvt_f32_s32_x2(svint32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svcvt_u32_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzu.x2.nxv4i32.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzu.x2.nxv4i32.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z21test_svcvt_u32_f32_x213svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzu.x2.nxv4i32.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzu.x2.nxv4i32.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svuint32x2_t test_svcvt_u32_f32_x2(svfloat32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvt_u32,_f32_x2,,)(zn);
@ -135,25 +133,27 @@ svuint32x2_t test_svcvt_u32_f32_x2(svfloat32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svcvt_s32_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzs.x2.nxv4i32.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzs.x2.nxv4i32.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z21test_svcvt_s32_f32_x213svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzs.x2.nxv4i32.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzs.x2.nxv4i32.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svint32x2_t test_svcvt_s32_f32_x2(svfloat32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvt_s32,_f32_x2,,)(zn);
@ -162,37 +162,35 @@ svint32x2_t test_svcvt_s32_f32_x2(svfloat32x2_t zn) __arm_streaming {
// x4
// CHECK-LABEL: @test_svcvt_f32_u32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ucvtf.x4.nxv4f32.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ucvtf.x4.nxv4f32.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z21test_svcvt_f32_u32_x412svuint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ucvtf.x4.nxv4f32.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ucvtf.x4.nxv4f32.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svcvt_f32_u32_x4(svuint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvt_f32,_u32_x4,,)(zn);
@ -200,37 +198,35 @@ svfloat32x4_t test_svcvt_f32_u32_x4(svuint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svcvt_f32_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.scvtf.x4.nxv4f32.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.scvtf.x4.nxv4f32.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z21test_svcvt_f32_s32_x411svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.scvtf.x4.nxv4f32.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.scvtf.x4.nxv4f32.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svcvt_f32_s32_x4(svint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvt_f32,_s32_x4,,)(zn);
@ -238,37 +234,35 @@ svfloat32x4_t test_svcvt_f32_s32_x4(svint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svcvt_u32_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzu.x4.nxv4i32.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzu.x4.nxv4i32.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z21test_svcvt_u32_f32_x413svfloat32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzu.x4.nxv4i32.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzu.x4.nxv4i32.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svuint32x4_t test_svcvt_u32_f32_x4(svfloat32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvt_u32,_f32_x4,,)(zn);
@ -276,37 +270,35 @@ svuint32x4_t test_svcvt_u32_f32_x4(svfloat32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svcvt_s32_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzs.x4.nxv4i32.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzs.x4.nxv4i32.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z21test_svcvt_s32_f32_x413svfloat32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzs.x4.nxv4i32.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.fcvtzs.x4.nxv4i32.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svint32x4_t test_svcvt_s32_f32_x4(svfloat32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvt_s32,_f32_x4,,)(zn);
@ -314,17 +306,13 @@ svint32x4_t test_svcvt_s32_f32_x4(svfloat32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvt_s16_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z20test_qcvt_s16_s32_x211svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svint16_t test_qcvt_s16_s32_x2(svint32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvt_s16,_s32_x2,,)(zn);
@ -332,17 +320,13 @@ svint16_t test_qcvt_s16_s32_x2(svint32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvt_u16_u32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z20test_qcvt_u16_u32_x212svuint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_qcvt_u16_u32_x2(svuint32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvt_u16,_u32_x2,,)(zn);
@ -350,17 +334,13 @@ svuint16_t test_qcvt_u16_u32_x2(svuint32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvt_u16_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z20test_qcvt_u16_s32_x211svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_qcvt_u16_s32_x2(svint32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvt_u16,_s32_x2,,)(zn);
@ -368,21 +348,13 @@ svuint16_t test_qcvt_u16_s32_x2(svint32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvt_u8_u32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqcvt.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqcvt.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z19test_qcvt_u8_u32_x412svuint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqcvt.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqcvt.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svuint8_t test_qcvt_u8_u32_x4(svuint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvt_u8,_u32_x4,,)(zn);
@ -390,21 +362,13 @@ svuint8_t test_qcvt_u8_u32_x4(svuint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvt_u16_u64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z20test_qcvt_u16_u64_x412svuint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_qcvt_u16_u64_x4(svuint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvt_u16,_u64_x4,,)(zn);
@ -412,21 +376,13 @@ svuint16_t test_qcvt_u16_u64_x4(svuint64x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvt_s8_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvt.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvt.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z19test_qcvt_s8_s32_x411svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvt.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvt.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svint8_t test_qcvt_s8_s32_x4(svint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvt_s8,_s32_x4,,)(zn);
@ -434,21 +390,13 @@ svint8_t test_qcvt_s8_s32_x4(svint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvt_s16_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z20test_qcvt_s16_s64_x411svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svint16_t test_qcvt_s16_s64_x4(svint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvt_s16,_s64_x4,,)(zn);
@ -456,21 +404,13 @@ svint16_t test_qcvt_s16_s64_x4(svint64x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvt_u8_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtu.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtu.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z19test_qcvt_u8_s32_x411svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtu.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtu.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svuint8_t test_qcvt_u8_s32_x4(svint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvt_u8,_s32_x4,,)(zn);
@ -478,21 +418,13 @@ svuint8_t test_qcvt_u8_s32_x4(svint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvt_u16_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z20test_qcvt_u16_s64_x411svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_qcvt_u16_s64_x4(svint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvt_u16,_s64_x4,,)(zn);
@ -500,21 +432,27 @@ svuint16_t test_qcvt_u16_s64_x4(svint64x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_cvt_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fcvt.widen.x2.nxv4f32(<vscale x 8 x half> [[ZN:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP4]]
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z15test_cvt_f32_x2u13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fcvt.widen.x2.nxv4f32(<vscale x 8 x half> [[ZN:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
__attribute__((target("sme-f16f16"))) svfloat32x2_t test_cvt_f32_x2(svfloat16_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvt_f32,_f16_x2,,)(zn);

10
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_cvtl.c
View File

@ -19,21 +19,27 @@
// CHECK-LABEL: @test_cvtl_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fcvtl.widen.x2.nxv4f32(<vscale x 8 x half> [[ZN:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP4]]
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z16test_cvtl_f32_x2u13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fcvtl.widen.x2.nxv4f32(<vscale x 8 x half> [[ZN:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_cvtl_f32_x2(svfloat16_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvtl_f32,_f16_x2,,)(zn);

120
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_cvtn.c
View File

@ -19,17 +19,13 @@
// CHECK-LABEL: @test_cvtn_f16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvtn.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvtn.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z16test_cvtn_f16_x213svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvtn.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fcvtn.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
//
svfloat16_t test_cvtn_f16_x2(svfloat32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvtn_f16,_f32_x2,,)(zn);
@ -37,17 +33,13 @@ svfloat16_t test_cvtn_f16_x2(svfloat32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_cvtn_bf16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.bfcvtn.x2(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.bfcvtn.x2(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z17test_cvtn_bf16_x213svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.bfcvtn.x2(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.bfcvtn.x2(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
//
svbfloat16_t test_cvtn_bf16_x2(svfloat32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svcvtn_bf16,_f32_x2,,)(zn);
@ -55,21 +47,13 @@ svbfloat16_t test_cvtn_bf16_x2(svfloat32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvtn_u8_u32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqcvtn.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqcvtn.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z20test_qcvtn_u8_u32_x412svuint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqcvtn.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqcvtn.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svuint8_t test_qcvtn_u8_u32_x4(svuint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvtn_u8,_u32_x4,,)(zn);
@ -77,21 +61,13 @@ svuint8_t test_qcvtn_u8_u32_x4(svuint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvtn_u16_u64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvtn.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvtn.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z21test_qcvtn_u16_u64_x412svuint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvtn.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvtn.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_qcvtn_u16_u64_x4(svuint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvtn_u16,_u64_x4,,)(zn);
@ -99,21 +75,13 @@ svuint16_t test_qcvtn_u16_u64_x4(svuint64x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvtn_s8_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtn.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtn.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z20test_qcvtn_s8_s32_x411svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtn.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtn.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svint8_t test_qcvtn_s8_s32_x4(svint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvtn_s8,_s32_x4,,)(zn);
@ -121,21 +89,13 @@ svint8_t test_qcvtn_s8_s32_x4(svint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvtn_s16_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtn.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtn.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z21test_qcvtn_s16_s64_x411svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtn.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtn.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svint16_t test_qcvtn_s16_s64_x4(svint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvtn_s16,_s64_x4,,)(zn);
@ -143,21 +103,13 @@ svint16_t test_qcvtn_s16_s64_x4(svint64x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvtn_u8_32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtun.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtun.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z19test_qcvtn_u8_32_x411svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtun.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtun.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svuint8_t test_qcvtn_u8_32_x4(svint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvtn_u8,_s32_x4,,)(zn);
@ -165,21 +117,13 @@ svuint8_t test_qcvtn_u8_32_x4(svint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_qcvtn_u16_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtun.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtun.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z21test_qcvtn_u16_s64_x411svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtun.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtun.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_qcvtn_u16_s64_x4(svint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqcvtn_u16,_s64_x4,,)(zn);

600
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_faminmax.c
View File

@ -19,29 +19,27 @@
// CHECK-LABEL: @test_svamax_f16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svamax_f16_x213svfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
svfloat16x2_t test_svamax_f16_x2(svfloat16x2_t zdn, svfloat16x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamax,_f16_x2)(zdn, zm);
@ -49,29 +47,27 @@ svfloat16x2_t test_svamax_f16_x2(svfloat16x2_t zdn, svfloat16x2_t zm) __arm_stre
// CHECK-LABEL: @test_svamax_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svamax_f32_x213svfloat32x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svamax_f32_x2(svfloat32x2_t zdn, svfloat32x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamax,_f32_x2)(zdn, zm);
@ -79,29 +75,27 @@ svfloat32x2_t test_svamax_f32_x2(svfloat32x2_t zdn, svfloat32x2_t zm) __arm_stre
// CHECK-LABEL: @test_svamax_f64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famax.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x double> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famax.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svamax_f64_x213svfloat64x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famax.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x double> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famax.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
svfloat64x2_t test_svamax_f64_x2(svfloat64x2_t zdn, svfloat64x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamax,_f64_x2)(zdn, zm);
@ -109,29 +103,27 @@ svfloat64x2_t test_svamax_f64_x2(svfloat64x2_t zdn, svfloat64x2_t zm) __arm_stre
// CHECK-LABEL: @test_svamin_f16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famin.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famin.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svamin_f16_x213svfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famin.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famin.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
svfloat16x2_t test_svamin_f16_x2(svfloat16x2_t zdn, svfloat16x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamin,_f16_x2)(zdn, zm);
@ -139,29 +131,27 @@ svfloat16x2_t test_svamin_f16_x2(svfloat16x2_t zdn, svfloat16x2_t zm) __arm_stre
// CHECK-LABEL: @test_svamin_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famin.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famin.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svamin_f32_x213svfloat32x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famin.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famin.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svamin_f32_x2(svfloat32x2_t zdn, svfloat32x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamin,_f32_x2)(zdn, zm);
@ -169,29 +159,27 @@ svfloat32x2_t test_svamin_f32_x2(svfloat32x2_t zdn, svfloat32x2_t zm) __arm_stre
// CHECK-LABEL: @test_svamin_f64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famin.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x double> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famin.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svamin_f64_x213svfloat64x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famin.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x double> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famin.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
svfloat64x2_t test_svamin_f64_x2(svfloat64x2_t zdn, svfloat64x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamin,_f64_x2)(zdn, zm);
@ -201,45 +189,35 @@ svfloat64x2_t test_svamin_f64_x2(svfloat64x2_t zdn, svfloat64x2_t zm) __arm_stre
// CHECK-LABEL: @test_svamax_f16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 8)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP12]], <vscale x 8 x half> [[TMP13]], i64 16)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP14]], <vscale x 8 x half> [[TMP15]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svamax_f16_x413svfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 8)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP12]], <vscale x 8 x half> [[TMP13]], i64 16)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP14]], <vscale x 8 x half> [[TMP15]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
svfloat16x4_t test_svamax_f16_x4(svfloat16x4_t zdn, svfloat16x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamax,_f16_x4)(zdn, zm);
@ -247,45 +225,35 @@ svfloat16x4_t test_svamax_f16_x4(svfloat16x4_t zdn, svfloat16x4_t zm) __arm_stre
// CHECK-LABEL: @test_svamax_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 4)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP12]], <vscale x 4 x float> [[TMP13]], i64 8)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP14]], <vscale x 4 x float> [[TMP15]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svamax_f32_x413svfloat32x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 4)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP12]], <vscale x 4 x float> [[TMP13]], i64 8)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP14]], <vscale x 4 x float> [[TMP15]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svamax_f32_x4(svfloat32x4_t zdn, svfloat32x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamax,_f32_x4)(zdn, zm);
@ -293,45 +261,35 @@ svfloat32x4_t test_svamax_f32_x4(svfloat32x4_t zdn, svfloat32x4_t zm) __arm_stre
// CHECK-LABEL: @test_svamax_f64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famax.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 2)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP12]], <vscale x 2 x double> [[TMP13]], i64 4)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP14]], <vscale x 2 x double> [[TMP15]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x double> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famax.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svamax_f64_x413svfloat64x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famax.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 2)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP12]], <vscale x 2 x double> [[TMP13]], i64 4)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP14]], <vscale x 2 x double> [[TMP15]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x double> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famax.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
svfloat64x4_t test_svamax_f64_x4(svfloat64x4_t zdn, svfloat64x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamax,_f64_x4)(zdn, zm);
@ -339,45 +297,35 @@ svfloat64x4_t test_svamax_f64_x4(svfloat64x4_t zdn, svfloat64x4_t zm) __arm_stre
// CHECK-LABEL: @test_svamin_f16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famin.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 8)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP12]], <vscale x 8 x half> [[TMP13]], i64 16)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP14]], <vscale x 8 x half> [[TMP15]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famin.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svamin_f16_x413svfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famin.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 8)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP12]], <vscale x 8 x half> [[TMP13]], i64 16)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP14]], <vscale x 8 x half> [[TMP15]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famin.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
svfloat16x4_t test_svamin_f16_x4(svfloat16x4_t zdn, svfloat16x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamin,_f16_x4)(zdn, zm);
@ -385,45 +333,35 @@ svfloat16x4_t test_svamin_f16_x4(svfloat16x4_t zdn, svfloat16x4_t zm) __arm_stre
// CHECK-LABEL: @test_svamin_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famin.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 4)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP12]], <vscale x 4 x float> [[TMP13]], i64 8)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP14]], <vscale x 4 x float> [[TMP15]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famin.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svamin_f32_x413svfloat32x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famin.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 4)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP12]], <vscale x 4 x float> [[TMP13]], i64 8)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP14]], <vscale x 4 x float> [[TMP15]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famin.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svamin_f32_x4(svfloat32x4_t zdn, svfloat32x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamin,_f32_x4)(zdn, zm);
@ -431,45 +369,35 @@ svfloat32x4_t test_svamin_f32_x4(svfloat32x4_t zdn, svfloat32x4_t zm) __arm_stre
// CHECK-LABEL: @test_svamin_f64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famin.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 2)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP12]], <vscale x 2 x double> [[TMP13]], i64 4)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP14]], <vscale x 2 x double> [[TMP15]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x double> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famin.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svamin_f64_x413svfloat64x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famin.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 2)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP12]], <vscale x 2 x double> [[TMP13]], i64 4)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP14]], <vscale x 2 x double> [[TMP15]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x double> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famin.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
svfloat64x4_t test_svamin_f64_x4(svfloat64x4_t zdn, svfloat64x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svamin,_f64_x4)(zdn, zm);

384
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_fmlas16.c
View File

@ -16,19 +16,15 @@
#endif
// CHECK-LABEL: define dso_local void @test_svmla_single_za16_f16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z32test_svmla_single_za16_f16_vg1x2j13svfloat16x2_tu13__SVFloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM]])
// CHECK-CXX-NEXT: ret void
//
void test_svmla_single_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -36,23 +32,15 @@ void test_svmla_single_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat1
}
// CHECK-LABEL: define dso_local void @test_svmla_single_za16_f16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z32test_svmla_single_za16_f16_vg1x4j13svfloat16x4_tu13__SVFloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM]])
// CHECK-CXX-NEXT: ret void
//
void test_svmla_single_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -60,19 +48,15 @@ void test_svmla_single_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat1
}
// CHECK-LABEL: define dso_local void @test_svmls_single_za16_f16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z32test_svmls_single_za16_f16_vg1x2j13svfloat16x2_tu13__SVFloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM]])
// CHECK-CXX-NEXT: ret void
//
void test_svmls_single_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -80,23 +64,15 @@ void test_svmls_single_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat1
}
// CHECK-LABEL: define dso_local void @test_svmls_single_za16_f16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z32test_svmls_single_za16_f16_vg1x4j13svfloat16x4_tu13__SVFloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM]])
// CHECK-CXX-NEXT: ret void
//
void test_svmls_single_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -104,23 +80,15 @@ void test_svmls_single_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat1
}
// CHECK-LABEL: define dso_local void @test_svmla_za16_f16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 16 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM_COERCE0]], <vscale x 8 x half> [[ZM_COERCE1]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z25test_svmla_za16_f16_vg1x2j13svfloat16x2_tS_(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 16 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 0)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM_COERCE0]], <vscale x 8 x half> [[ZM_COERCE1]])
// CHECK-CXX-NEXT: ret void
//
void test_svmla_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16x2_t zm) __arm_streaming __arm_inout("za") {
@ -128,31 +96,15 @@ void test_svmla_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16x2_t z
}
// CHECK-LABEL: define dso_local void @test_svmla_za16_f16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 32 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM_COERCE0]], <vscale x 8 x half> [[ZM_COERCE1]], <vscale x 8 x half> [[ZM_COERCE2]], <vscale x 8 x half> [[ZM_COERCE3]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z25test_svmla_za16_f16_vg1x4j13svfloat16x4_tS_(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 32 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-CXX-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 0)
// CHECK-CXX-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CHECK-CXX-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CHECK-CXX-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM_COERCE0]], <vscale x 8 x half> [[ZM_COERCE1]], <vscale x 8 x half> [[ZM_COERCE2]], <vscale x 8 x half> [[ZM_COERCE3]])
// CHECK-CXX-NEXT: ret void
//
void test_svmla_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16x4_t zm) __arm_streaming __arm_inout("za") {
@ -160,23 +112,15 @@ void test_svmla_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16x4_t z
}
// CHECK-LABEL: define dso_local void @test_svmls_za16_f16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 16 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM_COERCE0]], <vscale x 8 x half> [[ZM_COERCE1]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z25test_svmls_za16_f16_vg1x2j13svfloat16x2_tS_(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 16 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 0)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM_COERCE0]], <vscale x 8 x half> [[ZM_COERCE1]])
// CHECK-CXX-NEXT: ret void
//
void test_svmls_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16x2_t zm) __arm_streaming __arm_inout("za") {
@ -184,31 +128,15 @@ void test_svmls_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16x2_t z
}
// CHECK-LABEL: define dso_local void @test_svmls_za16_f16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 32 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM_COERCE0]], <vscale x 8 x half> [[ZM_COERCE1]], <vscale x 8 x half> [[ZM_COERCE2]], <vscale x 8 x half> [[ZM_COERCE3]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z25test_svmls_za16_f16_vg1x4j13svfloat16x4_tS_(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 32 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-CXX-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 0)
// CHECK-CXX-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CHECK-CXX-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CHECK-CXX-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM_COERCE0]], <vscale x 8 x half> [[ZM_COERCE1]], <vscale x 8 x half> [[ZM_COERCE2]], <vscale x 8 x half> [[ZM_COERCE3]])
// CHECK-CXX-NEXT: ret void
//
void test_svmls_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16x4_t zm) __arm_streaming __arm_inout("za") {
@ -216,19 +144,15 @@ void test_svmls_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16x4_t z
}
// CHECK-LABEL: define dso_local void @test_svmla_lane_za16_f16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z30test_svmla_lane_za16_f16_vg1x2j13svfloat16x2_tu13__SVFloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-CXX-NEXT: ret void
//
void test_svmla_lane_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -236,23 +160,15 @@ void test_svmla_lane_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16_
}
// CHECK-LABEL: define dso_local void @test_svmla_lane_za16_f16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z30test_svmla_lane_za16_f16_vg1x4j13svfloat16x4_tu13__SVFloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-CXX-NEXT: ret void
//
void test_svmla_lane_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -260,19 +176,15 @@ void test_svmla_lane_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16_
}
// CHECK-LABEL: define dso_local void @test_svmls_lane_za16_f16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z30test_svmls_lane_za16_f16_vg1x2j13svfloat16x2_tu13__SVFloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-CXX-NEXT: ret void
//
void test_svmls_lane_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -280,23 +192,15 @@ void test_svmls_lane_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16_
}
// CHECK-LABEL: define dso_local void @test_svmls_lane_za16_f16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z30test_svmls_lane_za16_f16_vg1x4j13svfloat16x4_tu13__SVFloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[ZN_COERCE0]], <vscale x 8 x half> [[ZN_COERCE1]], <vscale x 8 x half> [[ZN_COERCE2]], <vscale x 8 x half> [[ZN_COERCE3]], <vscale x 8 x half> [[ZM]], i32 7)
// CHECK-CXX-NEXT: ret void
//
void test_svmls_lane_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -304,19 +208,15 @@ void test_svmls_lane_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16_
}
// CHECK-LABEL: define dso_local void @test_svmla_single_za16_bf16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z33test_svmla_single_za16_bf16_vg1x2j14svbfloat16x2_tu14__SVBfloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-CXX-NEXT: ret void
//
void test_svmla_single_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -324,23 +224,15 @@ void test_svmla_single_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbflo
}
// CHECK-LABEL: define dso_local void @test_svmla_single_za16_bf16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z33test_svmla_single_za16_bf16_vg1x4j14svbfloat16x4_tu14__SVBfloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-CXX-NEXT: ret void
//
void test_svmla_single_za16_bf16_vg1x4(uint32_t slice, svbfloat16x4_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -348,19 +240,15 @@ void test_svmla_single_za16_bf16_vg1x4(uint32_t slice, svbfloat16x4_t zn, svbflo
}
// CHECK-LABEL: define dso_local void @test_svmls_single_za16_bf16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z33test_svmls_single_za16_bf16_vg1x2j14svbfloat16x2_tu14__SVBfloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-CXX-NEXT: ret void
//
void test_svmls_single_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -368,23 +256,15 @@ void test_svmls_single_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbflo
}
// CHECK-LABEL: define dso_local void @test_svmls_single_za16_bf16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z33test_svmls_single_za16_bf16_vg1x4j14svbfloat16x4_tu14__SVBfloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM]])
// CHECK-CXX-NEXT: ret void
//
void test_svmls_single_za16_bf16_vg1x4(uint32_t slice, svbfloat16x4_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -392,23 +272,15 @@ void test_svmls_single_za16_bf16_vg1x4(uint32_t slice, svbfloat16x4_t zn, svbflo
}
// CHECK-LABEL: define dso_local void @test_svmla_za16_bf16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 16 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM_COERCE0]], <vscale x 8 x bfloat> [[ZM_COERCE1]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z26test_svmla_za16_bf16_vg1x2j14svbfloat16x2_tS_(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 16 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 0)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM_COERCE0]], <vscale x 8 x bfloat> [[ZM_COERCE1]])
// CHECK-CXX-NEXT: ret void
//
void test_svmla_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbfloat16x2_t zm) __arm_streaming __arm_inout("za") {
@ -416,31 +288,15 @@ void test_svmla_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbfloat16x2_
}
// CHECK-LABEL: define dso_local void @test_svmla_za16_bf16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 32 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM_COERCE0]], <vscale x 8 x bfloat> [[ZM_COERCE1]], <vscale x 8 x bfloat> [[ZM_COERCE2]], <vscale x 8 x bfloat> [[ZM_COERCE3]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z26test_svmla_za16_bf16_vg1x4j14svbfloat16x4_tS_(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 32 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-CXX-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 0)
// CHECK-CXX-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CHECK-CXX-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CHECK-CXX-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM_COERCE0]], <vscale x 8 x bfloat> [[ZM_COERCE1]], <vscale x 8 x bfloat> [[ZM_COERCE2]], <vscale x 8 x bfloat> [[ZM_COERCE3]])
// CHECK-CXX-NEXT: ret void
//
void test_svmla_za16_bf16_vg1x4(uint32_t slice, svbfloat16x4_t zn, svbfloat16x4_t zm) __arm_streaming __arm_inout("za") {
@ -448,23 +304,15 @@ void test_svmla_za16_bf16_vg1x4(uint32_t slice, svbfloat16x4_t zn, svbfloat16x4_
}
// CHECK-LABEL: define dso_local void @test_svmls_za16_bf16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 16 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM_COERCE0]], <vscale x 8 x bfloat> [[ZM_COERCE1]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z26test_svmls_za16_bf16_vg1x2j14svbfloat16x2_tS_(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 16 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 0)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM_COERCE0]], <vscale x 8 x bfloat> [[ZM_COERCE1]])
// CHECK-CXX-NEXT: ret void
//
void test_svmls_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbfloat16x2_t zm) __arm_streaming __arm_inout("za") {
@ -472,31 +320,15 @@ void test_svmls_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbfloat16x2_
}
// CHECK-LABEL: define dso_local void @test_svmls_za16_bf16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 32 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM_COERCE0]], <vscale x 8 x bfloat> [[ZM_COERCE1]], <vscale x 8 x bfloat> [[ZM_COERCE2]], <vscale x 8 x bfloat> [[ZM_COERCE3]])
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z26test_svmls_za16_bf16_vg1x4j14svbfloat16x4_tS_(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 32 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-CXX-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 0)
// CHECK-CXX-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CHECK-CXX-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CHECK-CXX-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM_COERCE0]], <vscale x 8 x bfloat> [[ZM_COERCE1]], <vscale x 8 x bfloat> [[ZM_COERCE2]], <vscale x 8 x bfloat> [[ZM_COERCE3]])
// CHECK-CXX-NEXT: ret void
//
void test_svmls_za16_bf16_vg1x4(uint32_t slice, svbfloat16x4_t zn, svbfloat16x4_t zm) __arm_streaming __arm_inout("za") {
@ -504,19 +336,15 @@ void test_svmls_za16_bf16_vg1x4(uint32_t slice, svbfloat16x4_t zn, svbfloat16x4_
}
// CHECK-LABEL: define dso_local void @test_svmla_lane_za16_bf16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z31test_svmla_lane_za16_bf16_vg1x2j14svbfloat16x2_tu14__SVBfloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-CXX-NEXT: ret void
//
void test_svmla_lane_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -524,23 +352,15 @@ void test_svmla_lane_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbfloat
}
// CHECK-LABEL: define dso_local void @test_svmla_lane_za16_bf16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z31test_svmla_lane_za16_bf16_vg1x4j14svbfloat16x4_tu14__SVBfloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-CXX-NEXT: ret void
//
void test_svmla_lane_za16_bf16_vg1x4(uint32_t slice, svbfloat16x4_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -548,19 +368,15 @@ void test_svmla_lane_za16_bf16_vg1x4(uint32_t slice, svbfloat16x4_t zn, svbfloat
}
// CHECK-LABEL: define dso_local void @test_svmls_lane_za16_bf16_vg1x2(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z31test_svmls_lane_za16_bf16_vg1x2j14svbfloat16x2_tu14__SVBfloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-CXX-NEXT: ret void
//
void test_svmls_lane_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -568,23 +384,15 @@ void test_svmls_lane_za16_bf16_vg1x2(uint32_t slice, svbfloat16x2_t zn, svbfloat
}
// CHECK-LABEL: define dso_local void @test_svmls_lane_za16_bf16_vg1x4(
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-NEXT: ret void
//
// CHECK-CXX-LABEL: define dso_local void @_Z31test_svmls_lane_za16_bf16_vg1x4j14svbfloat16x4_tu14__SVBfloat16_t(
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x bfloat> [[ZN:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
// CHECK-CXX-NEXT: entry:
// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 0)
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv8bf16(i32 [[SLICE]], <vscale x 8 x bfloat> [[ZN_COERCE0]], <vscale x 8 x bfloat> [[ZN_COERCE1]], <vscale x 8 x bfloat> [[ZN_COERCE2]], <vscale x 8 x bfloat> [[ZN_COERCE3]], <vscale x 8 x bfloat> [[ZM]], i32 7)
// CHECK-CXX-NEXT: ret void
//
void test_svmls_lane_za16_bf16_vg1x4(uint32_t slice, svbfloat16x4_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {

144
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_fp_dots.c
View File

@ -19,20 +19,12 @@
// Multi, multi (half)
// CHECK-LABEL: @test_svdot_multi_za32_vg1x2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z31test_svdot_multi_za32_vg1x2_f16j13svfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svdot_multi_za32_vg1x2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16x2_t zm) __arm_streaming __arm_inout("za") {
@ -41,28 +33,12 @@ void test_svdot_multi_za32_vg1x2_f16(uint32_t slice_base, svfloat16x2_t zn, svfl
// CHECK-LABEL: @test_svdot_multi_za32_vg1x4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z31test_svdot_multi_za32_vg1x4_f16j13svfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svdot_multi_za32_vg1x4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16x4_t zm) __arm_streaming __arm_inout("za") {
@ -74,16 +50,12 @@ void test_svdot_multi_za32_vg1x4_f16(uint32_t slice_base, svfloat16x4_t zn, svfl
// Multi, single (half)
// CHECK-LABEL: @test_svdot_single_za32_vg1x2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z32test_svdot_single_za32_vg1x2_f16j13svfloat16x2_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svdot_single_za32_vg1x2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -92,20 +64,12 @@ void test_svdot_single_za32_vg1x2_f16(uint32_t slice_base, svfloat16x2_t zn, svf
// CHECK-LABEL: @test_svdot_single_za32_vg1x4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z32test_svdot_single_za32_vg1x4_f16j13svfloat16x4_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svdot_single_za32_vg1x4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -117,16 +81,12 @@ void test_svdot_single_za32_vg1x4_f16(uint32_t slice_base, svfloat16x4_t zn, svf
// Multi, indexed (half)
// CHECK-LABEL: @test_svdot_lane_za32_vg1x2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z30test_svdot_lane_za32_vg1x2_f16j13svfloat16x2_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svdot_lane_za32_vg1x2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -135,20 +95,12 @@ void test_svdot_lane_za32_vg1x2_f16(uint32_t slice_base, svfloat16x2_t zn, svflo
// CHECK-LABEL: @test_svdot_lane_za32_vg1x4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z30test_svdot_lane_za32_vg1x4_f16j13svfloat16x4_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svdot_lane_za32_vg1x4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -160,20 +112,12 @@ void test_svdot_lane_za32_vg1x4_f16(uint32_t slice_base, svfloat16x4_t zn, svflo
// Multi, multi (bfloat)
// CHECK-LABEL: @test_svdot_multi_za32_vg1x2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z32test_svdot_multi_za32_vg1x2_bf16j14svbfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svdot_multi_za32_vg1x2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16x2_t zm) __arm_streaming __arm_inout("za") {
@ -182,28 +126,12 @@ void test_svdot_multi_za32_vg1x2_bf16(uint32_t slice_base, svbfloat16x2_t zn, sv
// CHECK-LABEL: @test_svdot_multi_za32_vg1x4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z32test_svdot_multi_za32_vg1x4_bf16j14svbfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svdot_multi_za32_vg1x4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16x4_t zm) __arm_streaming __arm_inout("za") {
@ -215,16 +143,12 @@ void test_svdot_multi_za32_vg1x4_bf16(uint32_t slice_base, svbfloat16x4_t zn, sv
// Multi, single (bfloat)
// CHECK-LABEL: @test_svdot_single_za32_vg1x2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z33test_svdot_single_za32_vg1x2_bf16j14svbfloat16x2_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svdot_single_za32_vg1x2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -233,20 +157,12 @@ void test_svdot_single_za32_vg1x2_bf16(uint32_t slice_base, svbfloat16x2_t zn, s
// CHECK-LABEL: @test_svdot_single_za32_vg1x4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z33test_svdot_single_za32_vg1x4_bf16j14svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.single.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svdot_single_za32_vg1x4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -258,16 +174,12 @@ void test_svdot_single_za32_vg1x4_bf16(uint32_t slice_base, svbfloat16x4_t zn, s
// Multi, indexed (bfloat)
// CHECK-LABEL: @test_svdot_lane_za32_vg1x2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z31test_svdot_lane_za32_vg1x2_bf16j14svbfloat16x2_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svdot_lane_za32_vg1x2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -276,20 +188,12 @@ void test_svdot_lane_za32_vg1x2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svb
// CHECK-LABEL: @test_svdot_lane_za32_vg1x4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z31test_svdot_lane_za32_vg1x4_bf16j14svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fdot.lane.za32.vg1x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svdot_lane_za32_vg1x4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {

352
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_frint.c
View File

@ -21,25 +21,27 @@
// CHECK-LABEL: @test_svfrinta_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frinta.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frinta.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z20test_svfrinta_f32_x213svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frinta.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frinta.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svfrinta_f32_x2(svfloat32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svrinta,_f32_x2)(zn);
@ -47,37 +49,35 @@ svfloat32x2_t test_svfrinta_f32_x2(svfloat32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svfrinta_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frinta.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frinta.x4.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z20test_svfrinta_f32_x413svfloat32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frinta.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frinta.x4.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svfrinta_f32_x4(svfloat32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svrinta,_f32_x4)(zn);
@ -87,25 +87,27 @@ svfloat32x4_t test_svfrinta_f32_x4(svfloat32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svfrintam_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintm.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintm.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z21test_svfrintam_f32_x213svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintm.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintm.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svfrintam_f32_x2(svfloat32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svrintm,_f32_x2)(zn);
@ -113,37 +115,35 @@ svfloat32x2_t test_svfrintam_f32_x2(svfloat32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svfrintm_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintm.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintm.x4.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z20test_svfrintm_f32_x413svfloat32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintm.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintm.x4.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svfrintm_f32_x4(svfloat32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svrintm,_f32_x4)(zn);
@ -153,25 +153,27 @@ svfloat32x4_t test_svfrintm_f32_x4(svfloat32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svfrintn_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintn.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintn.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z20test_svfrintn_f32_x213svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintn.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintn.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svfrintn_f32_x2(svfloat32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svrintn,_f32_x2)(zn);
@ -179,37 +181,35 @@ svfloat32x2_t test_svfrintn_f32_x2(svfloat32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svfrintn_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintn.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintn.x4.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z20test_svfrintn_f32_x413svfloat32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintn.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintn.x4.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svfrintn_f32_x4(svfloat32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svrintn,_f32_x4)(zn);
@ -219,25 +219,27 @@ svfloat32x4_t test_svfrintn_f32_x4(svfloat32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svfrintp_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintp.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintp.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z20test_svfrintp_f32_x213svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintp.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintp.x2.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svfrintp_f32_x2(svfloat32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svrintp,_f32_x2)(zn);
@ -245,37 +247,35 @@ svfloat32x2_t test_svfrintp_f32_x2(svfloat32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svfrintp_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintp.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintp.x4.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z20test_svfrintp_f32_x413svfloat32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintp.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.frintp.x4.nxv4f32(<vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svfrintp_f32_x4(svfloat32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svrintp,_f32_x4)(zn);

576
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_int_dots.c
View File

File diff suppressed because it is too large Load Diff

90
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti2_lane_zt_x2.c
View File

@ -10,21 +10,27 @@
// CHECK-LABEL: @test_svluti2_lane_zt_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z23test_svluti2_lane_zt_u8u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
svuint8x2_t test_svluti2_lane_zt_u8(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_u8_x2(0, zn, 7);
@ -33,21 +39,27 @@ svuint8x2_t test_svluti2_lane_zt_u8(svuint8_t zn) __arm_streaming __arm_in("zt0"
// CHECK-LABEL: @test_svluti2_lane_zt_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z23test_svluti2_lane_zt_s8u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
svint8x2_t test_svluti2_lane_zt_s8(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_s8_x2(0, zn, 7);
@ -55,21 +67,27 @@ svint8x2_t test_svluti2_lane_zt_s8(svuint8_t zn) __arm_streaming __arm_in("zt0")
// CHECK-LABEL: @test_svluti2_lane_zt_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_u16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svuint16x2_t test_svluti2_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_u16_x2(0, zn, 7);
@ -78,21 +96,27 @@ svuint16x2_t test_svluti2_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_in("zt
// CHECK-LABEL: @test_svluti2_lane_zt_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_s16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svint16x2_t test_svluti2_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_s16_x2(0, zn, 7);
@ -100,21 +124,27 @@ svint16x2_t test_svluti2_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_in("zt0
// CHECK-LABEL: @test_svluti2_lane_zt_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP4]]
// CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_f16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
svfloat16x2_t test_svluti2_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_f16_x2(0, zn, 7);
@ -122,21 +152,27 @@ svfloat16x2_t test_svluti2_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_in("z
// CHECK-LABEL: @test_svluti2_lane_zt_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP4]]
// CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z25test_svluti2_lane_zt_bf16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
svbfloat16x2_t test_svluti2_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_bf16_x2(0, zn, 7);
@ -144,21 +180,27 @@ svbfloat16x2_t test_svluti2_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_in(
// CHECK-LABEL: @test_svluti2_lane_zt_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_u32u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svuint32x2_t test_svluti2_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_u32_x2(0, zn, 7);
@ -166,21 +208,27 @@ svuint32x2_t test_svluti2_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_in("zt
// CHECK-LABEL: @test_svluti2_lane_zt_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_s32u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svint32x2_t test_svluti2_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_s32_x2(0, zn, 7);
@ -188,21 +236,27 @@ svint32x2_t test_svluti2_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_in("zt0
// CHECK-LABEL: @test_svluti2_lane_zt_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP4]]
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_f32u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.luti2.lane.zt.x2.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svluti2_lane_zt_f32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_f32_x2(0, zn, 7);

90
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti2_lane_zt_x4.c
View File

@ -10,6 +10,7 @@
// CHECK-LABEL: @test_svluti2_lane_zt_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
@ -19,10 +20,13 @@
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP8]]
// CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z23test_svluti2_lane_zt_u8u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
@ -32,7 +36,9 @@
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
svuint8x4_t test_svluti2_lane_zt_u8(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_u8_x4(0, zn, 3);
@ -41,6 +47,7 @@ svuint8x4_t test_svluti2_lane_zt_u8(svuint8_t zn) __arm_streaming __arm_in("zt0"
// CHECK-LABEL: @test_svluti2_lane_zt_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
@ -50,10 +57,13 @@ svuint8x4_t test_svluti2_lane_zt_u8(svuint8_t zn) __arm_streaming __arm_in("zt0"
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP8]]
// CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z23test_svluti2_lane_zt_s8u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
@ -63,7 +73,9 @@ svuint8x4_t test_svluti2_lane_zt_u8(svuint8_t zn) __arm_streaming __arm_in("zt0"
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
svint8x4_t test_svluti2_lane_zt_s8(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_s8_x4(0, zn, 3);
@ -71,6 +83,7 @@ svint8x4_t test_svluti2_lane_zt_s8(svuint8_t zn) __arm_streaming __arm_in("zt0")
// CHECK-LABEL: @test_svluti2_lane_zt_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
@ -80,10 +93,13 @@ svint8x4_t test_svluti2_lane_zt_s8(svuint8_t zn) __arm_streaming __arm_in("zt0")
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP8]]
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_u16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
@ -93,7 +109,9 @@ svint8x4_t test_svluti2_lane_zt_s8(svuint8_t zn) __arm_streaming __arm_in("zt0")
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svuint16x4_t test_svluti2_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_u16_x4(0, zn, 3);
@ -101,6 +119,7 @@ svuint16x4_t test_svluti2_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_in("zt
// CHECK-LABEL: @test_svluti2_lane_zt_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
@ -110,10 +129,13 @@ svuint16x4_t test_svluti2_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_in("zt
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP8]]
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_s16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
@ -123,7 +145,9 @@ svuint16x4_t test_svluti2_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_in("zt
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svint16x4_t test_svluti2_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_s16_x4(0, zn, 3);
@ -131,6 +155,7 @@ svint16x4_t test_svluti2_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_in("zt0
// CHECK-LABEL: @test_svluti2_lane_zt_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
@ -140,10 +165,13 @@ svint16x4_t test_svluti2_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_in("zt0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP8]]
// CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_f16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
@ -153,7 +181,9 @@ svint16x4_t test_svluti2_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_in("zt0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
svfloat16x4_t test_svluti2_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_f16_x4(0, zn, 3);
@ -161,6 +191,7 @@ svfloat16x4_t test_svluti2_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_in("z
// CHECK-LABEL: @test_svluti2_lane_zt_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
@ -170,10 +201,13 @@ svfloat16x4_t test_svluti2_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_in("z
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP8]]
// CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z25test_svluti2_lane_zt_bf16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
@ -183,7 +217,9 @@ svfloat16x4_t test_svluti2_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_in("z
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
svbfloat16x4_t test_svluti2_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_bf16_x4(0, zn, 3);
@ -191,6 +227,7 @@ svbfloat16x4_t test_svluti2_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_in(
// CHECK-LABEL: @test_svluti2_lane_zt_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
@ -200,10 +237,13 @@ svbfloat16x4_t test_svluti2_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_in(
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP8]]
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_u32u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
@ -213,7 +253,9 @@ svbfloat16x4_t test_svluti2_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_in(
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svuint32x4_t test_svluti2_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_u32_x4(0, zn, 3);
@ -221,6 +263,7 @@ svuint32x4_t test_svluti2_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_in("zt
// CHECK-LABEL: @test_svluti2_lane_zt_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
@ -230,10 +273,13 @@ svuint32x4_t test_svluti2_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_in("zt
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP8]]
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_s32u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
@ -243,7 +289,9 @@ svuint32x4_t test_svluti2_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_in("zt
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svint32x4_t test_svluti2_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_s32_x4(0, zn, 3);
@ -251,6 +299,7 @@ svint32x4_t test_svluti2_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_in("zt0
// CHECK-LABEL: @test_svluti2_lane_zt_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
@ -260,10 +309,13 @@ svint32x4_t test_svluti2_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_in("zt0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP8]]
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_f32u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.luti2.lane.zt.x4.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
@ -273,7 +325,9 @@ svint32x4_t test_svluti2_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_in("zt0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svluti2_lane_zt_f32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti2_lane_zt_f32_x4(0, zn, 3);

90
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti4_lane_zt_x2.c
View File

@ -10,21 +10,27 @@
// CHECK-LABEL: @test_svluti4_lane_zt_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z23test_svluti4_lane_zt_u8u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
svuint8x2_t test_svluti4_lane_zt_u8(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_u8_x2(0, zn, 3);
@ -33,21 +39,27 @@ svuint8x2_t test_svluti4_lane_zt_u8(svuint8_t zn) __arm_streaming __arm_in("zt0"
// CHECK-LABEL: @test_svluti4_lane_zt_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z23test_svluti4_lane_zt_s8u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
svint8x2_t test_svluti4_lane_zt_s8(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_s8_x2(0, zn, 3);
@ -55,21 +67,27 @@ svint8x2_t test_svluti4_lane_zt_s8(svuint8_t zn) __arm_streaming __arm_in("zt0")
// CHECK-LABEL: @test_svluti4_lane_zt_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_u16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svuint16x2_t test_svluti4_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_u16_x2(0, zn, 3);
@ -78,21 +96,27 @@ svuint16x2_t test_svluti4_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_in("zt
// CHECK-LABEL: @test_svluti4_lane_zt_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_s16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svint16x2_t test_svluti4_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_s16_x2(0, zn, 3);
@ -100,21 +124,27 @@ svint16x2_t test_svluti4_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_in("zt0
// CHECK-LABEL: @test_svluti4_lane_zt_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP4]]
// CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_f16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
svfloat16x2_t test_svluti4_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_f16_x2(0, zn, 3);
@ -122,21 +152,27 @@ svfloat16x2_t test_svluti4_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_in("z
// CHECK-LABEL: @test_svluti4_lane_zt_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP4]]
// CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z25test_svluti4_lane_zt_bf16u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
svbfloat16x2_t test_svluti4_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_bf16_x2(0, zn, 3);
@ -144,21 +180,27 @@ svbfloat16x2_t test_svluti4_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_in(
// CHECK-LABEL: @test_svluti4_lane_zt_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_u32u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svuint32x2_t test_svluti4_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_u32_x2(0, zn, 3);
@ -166,21 +208,27 @@ svuint32x2_t test_svluti4_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_in("zt
// CHECK-LABEL: @test_svluti4_lane_zt_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_s32u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svint32x2_t test_svluti4_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_s32_x2(0, zn, 3);
@ -188,21 +236,27 @@ svint32x2_t test_svluti4_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_in("zt0
// CHECK-LABEL: @test_svluti4_lane_zt_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP4]]
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_f32u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.luti4.lane.zt.x2.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 3)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svluti4_lane_zt_f32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_f32_x2(0, zn, 3);

98
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti4_lane_zt_x4.c
View File

@ -8,9 +8,10 @@
#include <arm_sme.h>
// CHECK-LABEL: define dso_local <vscale x 32 x i16> @test_svluti4_lane_zt_u16
// CHECK-LABEL: define dso_local { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @test_svluti4_lane_zt_u16
// CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
@ -20,11 +21,14 @@
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP8]]
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: define dso_local <vscale x 32 x i16> @_Z24test_svluti4_lane_zt_u16u11__SVUint8_t
// CPP-CHECK-LABEL: define dso_local { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @_Z24test_svluti4_lane_zt_u16u11__SVUint8_t
// CPP-CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0:[0-9]+]] {
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
@ -34,15 +38,18 @@
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svuint16x4_t test_svluti4_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_u16_x4(0, zn, 1);
}
// CHECK-LABEL: define dso_local <vscale x 32 x half> @test_svluti4_lane_zt_f16
// CHECK-LABEL: define dso_local { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @test_svluti4_lane_zt_f16
// CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
@ -52,11 +59,14 @@ svuint16x4_t test_svluti4_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_in("zt
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP8]]
// CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
// CPP-CHECK-LABEL: define dso_local <vscale x 32 x half> @_Z24test_svluti4_lane_zt_f16u11__SVUint8_t
// CPP-CHECK-LABEL: define dso_local { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @_Z24test_svluti4_lane_zt_f16u11__SVUint8_t
// CPP-CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
@ -66,15 +76,18 @@ svuint16x4_t test_svluti4_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_in("zt
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
svfloat16x4_t test_svluti4_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_f16_x4(0, zn, 1);
}
// CHECK-LABEL: define dso_local <vscale x 32 x bfloat> @test_svluti4_lane_zt_bf16
// CHECK-LABEL: define dso_local { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @test_svluti4_lane_zt_bf16
// CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
@ -84,11 +97,14 @@ svfloat16x4_t test_svluti4_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_in("z
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP8]]
// CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
// CPP-CHECK-LABEL: define dso_local <vscale x 32 x bfloat> @_Z25test_svluti4_lane_zt_bf16u11__SVUint8_t
// CPP-CHECK-LABEL: define dso_local { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @_Z25test_svluti4_lane_zt_bf16u11__SVUint8_t
// CPP-CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
@ -98,15 +114,18 @@ svfloat16x4_t test_svluti4_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_in("z
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
svbfloat16x4_t test_svluti4_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_bf16_x4(0, zn, 1);
}
// CHECK-LABEL: define dso_local <vscale x 32 x i16> @test_svluti4_lane_zt_s16
// CHECK-LABEL: define dso_local { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @test_svluti4_lane_zt_s16
// CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
@ -116,11 +135,14 @@ svbfloat16x4_t test_svluti4_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_in(
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP8]]
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: define dso_local <vscale x 32 x i16> @_Z24test_svluti4_lane_zt_s16u11__SVUint8_t
// CPP-CHECK-LABEL: define dso_local { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @_Z24test_svluti4_lane_zt_s16u11__SVUint8_t
// CPP-CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
@ -130,15 +152,18 @@ svbfloat16x4_t test_svluti4_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_in(
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svint16x4_t test_svluti4_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_s16_x4(0, zn, 1);
}
// CHECK-LABEL: define dso_local <vscale x 16 x i32> @test_svluti4_lane_zt_u32
// CHECK-LABEL: define dso_local { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @test_svluti4_lane_zt_u32
// CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
@ -148,11 +173,14 @@ svint16x4_t test_svluti4_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_in("zt0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP8]]
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: define dso_local <vscale x 16 x i32> @_Z24test_svluti4_lane_zt_u32u11__SVUint8_t
// CPP-CHECK-LABEL: define dso_local { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @_Z24test_svluti4_lane_zt_u32u11__SVUint8_t
// CPP-CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
@ -162,15 +190,18 @@ svint16x4_t test_svluti4_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_in("zt0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svuint32x4_t test_svluti4_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_u32_x4(0, zn, 1);
}
// CHECK-LABEL: define dso_local <vscale x 16 x i32> @test_svluti4_lane_zt_s32
// CHECK-LABEL: define dso_local { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @test_svluti4_lane_zt_s32
// CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
@ -180,11 +211,14 @@ svuint32x4_t test_svluti4_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_in("zt
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP8]]
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: define dso_local <vscale x 16 x i32> @_Z24test_svluti4_lane_zt_s32u11__SVUint8_t
// CPP-CHECK-LABEL: define dso_local { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @_Z24test_svluti4_lane_zt_s32u11__SVUint8_t
// CPP-CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
@ -194,15 +228,18 @@ svuint32x4_t test_svluti4_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_in("zt
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svint32x4_t test_svluti4_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_s32_x4(0, zn, 1);
}
// CHECK-LABEL: define dso_local <vscale x 16 x float> @test_svluti4_lane_zt_f32
// CHECK-LABEL: define dso_local { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @test_svluti4_lane_zt_f32
// CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
@ -212,11 +249,14 @@ svint32x4_t test_svluti4_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_in("zt0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP8]]
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: define dso_local <vscale x 16 x float> @_Z24test_svluti4_lane_zt_f32u11__SVUint8_t
// CPP-CHECK-LABEL: define dso_local { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @_Z24test_svluti4_lane_zt_f32u11__SVUint8_t
// CPP-CHECK-SAME: (<vscale x 16 x i8> [[ZN:%.*]]) #[[ATTR0]] {
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.luti4.lane.zt.x4.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN]], i32 1)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
@ -226,7 +266,9 @@ svint32x4_t test_svluti4_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_in("zt0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP8]]
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svluti4_lane_zt_f32(svuint8_t zn) __arm_streaming __arm_in("zt0") {
return svluti4_lane_zt_f32_x4(0, zn, 1);

2256
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_max.c
View File

File diff suppressed because it is too large Load Diff

752
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_maxnm.c
View File

@ -19,25 +19,27 @@
// CHECK-LABEL: @test_svmaxnm_single_bf16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z27test_svmaxnm_single_bf16_x214svbfloat16x2_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
svbfloat16x2_t test_svmaxnm_single_bf16_x2(svbfloat16x2_t zdn, svbfloat16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_single_bf16_x2)(zdn, zm);
@ -45,25 +47,27 @@ svbfloat16x2_t test_svmaxnm_single_bf16_x2(svbfloat16x2_t zdn, svbfloat16_t zm)
// CHECK-LABEL: @test_svmaxnm_single_f16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z26test_svmaxnm_single_f16_x213svfloat16x2_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
svfloat16x2_t test_svmaxnm_single_f16_x2(svfloat16x2_t zdn, svfloat16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_single_f16_x2)(zdn, zm);
@ -71,25 +75,27 @@ svfloat16x2_t test_svmaxnm_single_f16_x2(svfloat16x2_t zdn, svfloat16_t zm) __ar
// CHECK-LABEL: @test_svmaxnm_single_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z26test_svmaxnm_single_f32_x213svfloat32x2_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svmaxnm_single_f32_x2(svfloat32x2_t zdn, svfloat32_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_single_f32_x2)(zdn, zm);
@ -97,25 +103,27 @@ svfloat32x2_t test_svmaxnm_single_f32_x2(svfloat32x2_t zdn, svfloat32_t zm) __ar
// CHECK-LABEL: @test_svmaxnm_single_f64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x double> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z26test_svmaxnm_single_f64_x213svfloat64x2_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x double> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.single.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
svfloat64x2_t test_svmaxnm_single_f64_x2(svfloat64x2_t zdn, svfloat64_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_single_f64_x2)(zdn, zm);
@ -125,37 +133,35 @@ svfloat64x2_t test_svmaxnm_single_f64_x2(svfloat64x2_t zdn, svfloat64_t zm) __ar
// CHECK-LABEL: @test_svmaxnm_single_bf16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP8]], <vscale x 8 x bfloat> [[TMP9]], i64 16)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP10]], <vscale x 8 x bfloat> [[TMP11]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z27test_svmaxnm_single_bf16_x414svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP8]], <vscale x 8 x bfloat> [[TMP9]], i64 16)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP10]], <vscale x 8 x bfloat> [[TMP11]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
svbfloat16x4_t test_svmaxnm_single_bf16_x4(svbfloat16x4_t zdn, svbfloat16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_single_bf16_x4)(zdn, zm);
@ -163,37 +169,35 @@ svbfloat16x4_t test_svmaxnm_single_bf16_x4(svbfloat16x4_t zdn, svbfloat16_t zm)
// CHECK-LABEL: @test_svmaxnm_single_f16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP8]], <vscale x 8 x half> [[TMP9]], i64 16)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z26test_svmaxnm_single_f16_x413svfloat16x4_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP8]], <vscale x 8 x half> [[TMP9]], i64 16)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
svfloat16x4_t test_svmaxnm_single_f16_x4(svfloat16x4_t zdn, svfloat16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_single_f16_x4)(zdn, zm);
@ -201,37 +205,35 @@ svfloat16x4_t test_svmaxnm_single_f16_x4(svfloat16x4_t zdn, svfloat16_t zm) __ar
// CHECK-LABEL: @test_svmaxnm_single_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z26test_svmaxnm_single_f32_x413svfloat32x4_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svmaxnm_single_f32_x4(svfloat32x4_t zdn, svfloat32_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_single_f32_x4)(zdn, zm);
@ -239,37 +241,35 @@ svfloat32x4_t test_svmaxnm_single_f32_x4(svfloat32x4_t zdn, svfloat32_t zm) __ar
// CHECK-LABEL: @test_svmaxnm_single_f64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP8]], <vscale x 2 x double> [[TMP9]], i64 4)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x double> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z26test_svmaxnm_single_f64_x413svfloat64x4_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP8]], <vscale x 2 x double> [[TMP9]], i64 4)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x double> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.single.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
svfloat64x4_t test_svmaxnm_single_f64_x4(svfloat64x4_t zdn, svfloat64_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_single_f64_x4)(zdn, zm);
@ -279,29 +279,27 @@ svfloat64x4_t test_svmaxnm_single_f64_x4(svfloat64x4_t zdn, svfloat64_t zm) __ar
// CHECK-LABEL: @test_svmaxnm_multi_bf16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.x2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.x2.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z26test_svmaxnm_multi_bf16_x214svbfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.x2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.x2.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
svbfloat16x2_t test_svmaxnm_multi_bf16_x2(svbfloat16x2_t zdn, svbfloat16x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_bf16_x2)(zdn, zm);
@ -309,29 +307,27 @@ svbfloat16x2_t test_svmaxnm_multi_bf16_x2(svbfloat16x2_t zdn, svbfloat16x2_t zm)
// CHECK-LABEL: @test_svmaxnm_multi_f16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z25test_svmaxnm_multi_f16_x213svfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
svfloat16x2_t test_svmaxnm_multi_f16_x2(svfloat16x2_t zdn, svfloat16x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_f16_x2)(zdn, zm);
@ -339,29 +335,27 @@ svfloat16x2_t test_svmaxnm_multi_f16_x2(svfloat16x2_t zdn, svfloat16x2_t zm) __a
// CHECK-LABEL: @test_svmaxnm_multi_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z25test_svmaxnm_multi_f32_x213svfloat32x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svmaxnm_multi_f32_x2(svfloat32x2_t zdn, svfloat32x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_f32_x2)(zdn, zm);
@ -369,29 +363,27 @@ svfloat32x2_t test_svmaxnm_multi_f32_x2(svfloat32x2_t zdn, svfloat32x2_t zm) __a
// CHECK-LABEL: @test_svmaxnm_multi_f64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x double> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z25test_svmaxnm_multi_f64_x213svfloat64x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x double> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
svfloat64x2_t test_svmaxnm_multi_f64_x2(svfloat64x2_t zdn, svfloat64x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_f64_x2)(zdn, zm);
@ -401,45 +393,35 @@ svfloat64x2_t test_svmaxnm_multi_f64_x2(svfloat64x2_t zdn, svfloat64x2_t zm) __a
// CHECK-LABEL: @test_svmaxnm_multi_bf16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.x4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP10]], <vscale x 8 x bfloat> [[TMP11]], i64 8)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP12]], <vscale x 8 x bfloat> [[TMP13]], i64 16)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP14]], <vscale x 8 x bfloat> [[TMP15]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.x4.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z26test_svmaxnm_multi_bf16_x414svbfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.x4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP10]], <vscale x 8 x bfloat> [[TMP11]], i64 8)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP12]], <vscale x 8 x bfloat> [[TMP13]], i64 16)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP14]], <vscale x 8 x bfloat> [[TMP15]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fmaxnm.x4.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
svbfloat16x4_t test_svmaxnm_multi_bf16_x4(svbfloat16x4_t zdn, svbfloat16x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_bf16_x4)(zdn, zm);
@ -447,45 +429,35 @@ svbfloat16x4_t test_svmaxnm_multi_bf16_x4(svbfloat16x4_t zdn, svbfloat16x4_t zm)
// CHECK-LABEL: @test_svmaxnm_multi_f16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 8)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP12]], <vscale x 8 x half> [[TMP13]], i64 16)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP14]], <vscale x 8 x half> [[TMP15]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z25test_svmaxnm_multi_f16_x413svfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 8)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP12]], <vscale x 8 x half> [[TMP13]], i64 16)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP14]], <vscale x 8 x half> [[TMP15]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fmaxnm.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
svfloat16x4_t test_svmaxnm_multi_f16_x4(svfloat16x4_t zdn, svfloat16x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_f16_x4)(zdn, zm);
@ -493,45 +465,35 @@ svfloat16x4_t test_svmaxnm_multi_f16_x4(svfloat16x4_t zdn, svfloat16x4_t zm) __a
// CHECK-LABEL: @test_svmaxnm_multi_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 4)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP12]], <vscale x 4 x float> [[TMP13]], i64 8)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP14]], <vscale x 4 x float> [[TMP15]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z25test_svmaxnm_multi_f32_x413svfloat32x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 4)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP12]], <vscale x 4 x float> [[TMP13]], i64 8)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP14]], <vscale x 4 x float> [[TMP15]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fmaxnm.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svmaxnm_multi_f32_x4(svfloat32x4_t zdn, svfloat32x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_f32_x4)(zdn, zm);
@ -539,45 +501,35 @@ svfloat32x4_t test_svmaxnm_multi_f32_x4(svfloat32x4_t zdn, svfloat32x4_t zm) __a
// CHECK-LABEL: @test_svmaxnm_multi_f64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 2)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP12]], <vscale x 2 x double> [[TMP13]], i64 4)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP14]], <vscale x 2 x double> [[TMP15]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x double> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z25test_svmaxnm_multi_f64_x413svfloat64x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 2)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP12]], <vscale x 2 x double> [[TMP13]], i64 4)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP14]], <vscale x 2 x double> [[TMP15]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x double> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fmaxnm.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
svfloat64x4_t test_svmaxnm_multi_f64_x4(svfloat64x4_t zdn, svfloat64x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svmaxnm,_f64_x4)(zdn, zm);

2256
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_min.c
View File

File diff suppressed because it is too large Load Diff

752
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_minnm.c
View File

@ -19,25 +19,27 @@
// CHECK-LABEL: @test_svminnm_single_bf16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.single.x2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.single.x2.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z27test_svminnm_single_bf16_x214svbfloat16x2_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.single.x2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.single.x2.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
svbfloat16x2_t test_svminnm_single_bf16_x2(svbfloat16x2_t zdn, svbfloat16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_single_bf16_x2)(zdn, zm);
@ -45,25 +47,27 @@ svbfloat16x2_t test_svminnm_single_bf16_x2(svbfloat16x2_t zdn, svbfloat16_t zm)
// CHECK-LABEL: @test_svminnm_single_f16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.single.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.single.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z26test_svminnm_single_f16_x213svfloat16x2_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.single.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.single.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
svfloat16x2_t test_svminnm_single_f16_x2(svfloat16x2_t zdn, svfloat16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_single_f16_x2)(zdn, zm);
@ -71,25 +75,27 @@ svfloat16x2_t test_svminnm_single_f16_x2(svfloat16x2_t zdn, svfloat16_t zm) __ar
// CHECK-LABEL: @test_svminnm_single_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.single.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.single.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z26test_svminnm_single_f32_x213svfloat32x2_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.single.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.single.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svminnm_single_f32_x2(svfloat32x2_t zdn, svfloat32_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_single_f32_x2)(zdn, zm);
@ -97,25 +103,27 @@ svfloat32x2_t test_svminnm_single_f32_x2(svfloat32x2_t zdn, svfloat32_t zm) __ar
// CHECK-LABEL: @test_svminnm_single_f64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.single.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x double> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.single.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z26test_svminnm_single_f64_x213svfloat64x2_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.single.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x double> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.single.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
svfloat64x2_t test_svminnm_single_f64_x2(svfloat64x2_t zdn, svfloat64_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_single_f64_x2)(zdn, zm);
@ -125,37 +133,35 @@ svfloat64x2_t test_svminnm_single_f64_x2(svfloat64x2_t zdn, svfloat64_t zm) __ar
// CHECK-LABEL: @test_svminnm_single_bf16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.single.x4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP8]], <vscale x 8 x bfloat> [[TMP9]], i64 16)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP10]], <vscale x 8 x bfloat> [[TMP11]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.single.x4.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z27test_svminnm_single_bf16_x414svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.single.x4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP8]], <vscale x 8 x bfloat> [[TMP9]], i64 16)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP10]], <vscale x 8 x bfloat> [[TMP11]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.single.x4.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
svbfloat16x4_t test_svminnm_single_bf16_x4(svbfloat16x4_t zdn, svbfloat16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_single_bf16_x4)(zdn, zm);
@ -163,37 +169,35 @@ svbfloat16x4_t test_svminnm_single_bf16_x4(svbfloat16x4_t zdn, svbfloat16_t zm)
// CHECK-LABEL: @test_svminnm_single_f16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.single.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP8]], <vscale x 8 x half> [[TMP9]], i64 16)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.single.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z26test_svminnm_single_f16_x413svfloat16x4_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.single.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP8]], <vscale x 8 x half> [[TMP9]], i64 16)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.single.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
svfloat16x4_t test_svminnm_single_f16_x4(svfloat16x4_t zdn, svfloat16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_single_f16_x4)(zdn, zm);
@ -201,37 +205,35 @@ svfloat16x4_t test_svminnm_single_f16_x4(svfloat16x4_t zdn, svfloat16_t zm) __ar
// CHECK-LABEL: @test_svminnm_single_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.single.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.single.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z26test_svminnm_single_f32_x413svfloat32x4_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.single.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.single.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svminnm_single_f32_x4(svfloat32x4_t zdn, svfloat32_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_single_f32_x4)(zdn, zm);
@ -239,37 +241,35 @@ svfloat32x4_t test_svminnm_single_f32_x4(svfloat32x4_t zdn, svfloat32_t zm) __ar
// CHECK-LABEL: @test_svminnm_single_f64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.single.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP8]], <vscale x 2 x double> [[TMP9]], i64 4)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x double> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.single.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z26test_svminnm_single_f64_x413svfloat64x4_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.single.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP8]], <vscale x 2 x double> [[TMP9]], i64 4)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x double> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.single.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
svfloat64x4_t test_svminnm_single_f64_x4(svfloat64x4_t zdn, svfloat64_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_single_f64_x4)(zdn, zm);
@ -279,29 +279,27 @@ svfloat64x4_t test_svminnm_single_f64_x4(svfloat64x4_t zdn, svfloat64_t zm) __ar
// CHECK-LABEL: @test_svminnm_multi_bf16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.x2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.x2.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z26test_svminnm_multi_bf16_x214svbfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.x2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.x2.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
svbfloat16x2_t test_svminnm_multi_bf16_x2(svbfloat16x2_t zdn, svbfloat16x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_bf16_x2)(zdn, zm);
@ -309,29 +307,27 @@ svbfloat16x2_t test_svminnm_multi_bf16_x2(svbfloat16x2_t zdn, svbfloat16x2_t zm)
// CHECK-LABEL: @test_svminnm_multi_f16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z25test_svminnm_multi_f16_x213svfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.x2.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
svfloat16x2_t test_svminnm_multi_f16_x2(svfloat16x2_t zdn, svfloat16x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_f16_x2)(zdn, zm);
@ -339,29 +335,27 @@ svfloat16x2_t test_svminnm_multi_f16_x2(svfloat16x2_t zdn, svfloat16x2_t zm) __a
// CHECK-LABEL: @test_svminnm_multi_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z25test_svminnm_multi_f32_x213svfloat32x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.x2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.x2.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svminnm_multi_f32_x2(svfloat32x2_t zdn, svfloat32x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_f32_x2)(zdn, zm);
@ -369,29 +363,27 @@ svfloat32x2_t test_svminnm_multi_f32_x2(svfloat32x2_t zdn, svfloat32x2_t zm) __a
// CHECK-LABEL: @test_svminnm_multi_f64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x double> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z25test_svminnm_multi_f64_x213svfloat64x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.x2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x double> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.x2.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
svfloat64x2_t test_svminnm_multi_f64_x2(svfloat64x2_t zdn, svfloat64x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_f64_x2)(zdn, zm);
@ -401,45 +393,35 @@ svfloat64x2_t test_svminnm_multi_f64_x2(svfloat64x2_t zdn, svfloat64x2_t zm) __a
// CHECK-LABEL: @test_svminnm_multi_bf16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.x4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP10]], <vscale x 8 x bfloat> [[TMP11]], i64 8)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP12]], <vscale x 8 x bfloat> [[TMP13]], i64 16)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP14]], <vscale x 8 x bfloat> [[TMP15]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.x4.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z26test_svminnm_multi_bf16_x414svbfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.x4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP10]], <vscale x 8 x bfloat> [[TMP11]], i64 8)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP12]], <vscale x 8 x bfloat> [[TMP13]], i64 16)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP14]], <vscale x 8 x bfloat> [[TMP15]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.fminnm.x4.nxv8bf16(<vscale x 8 x bfloat> [[ZDN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZDN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
svbfloat16x4_t test_svminnm_multi_bf16_x4(svbfloat16x4_t zdn, svbfloat16x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_bf16_x4)(zdn, zm);
@ -447,45 +429,35 @@ svbfloat16x4_t test_svminnm_multi_bf16_x4(svbfloat16x4_t zdn, svbfloat16x4_t zm)
// CHECK-LABEL: @test_svminnm_multi_f16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 8)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP12]], <vscale x 8 x half> [[TMP13]], i64 16)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP14]], <vscale x 8 x half> [[TMP15]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z25test_svminnm_multi_f16_x413svfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.x4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 8)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP12]], <vscale x 8 x half> [[TMP13]], i64 16)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP14]], <vscale x 8 x half> [[TMP15]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.fminnm.x4.nxv8f16(<vscale x 8 x half> [[ZDN_COERCE0:%.*]], <vscale x 8 x half> [[ZDN_COERCE1:%.*]], <vscale x 8 x half> [[ZDN_COERCE2:%.*]], <vscale x 8 x half> [[ZDN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
svfloat16x4_t test_svminnm_multi_f16_x4(svfloat16x4_t zdn, svfloat16x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_f16_x4)(zdn, zm);
@ -493,45 +465,35 @@ svfloat16x4_t test_svminnm_multi_f16_x4(svfloat16x4_t zdn, svfloat16x4_t zm) __a
// CHECK-LABEL: @test_svminnm_multi_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 4)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP12]], <vscale x 4 x float> [[TMP13]], i64 8)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP14]], <vscale x 4 x float> [[TMP15]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z25test_svminnm_multi_f32_x413svfloat32x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZDN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.x4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 4)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP12]], <vscale x 4 x float> [[TMP13]], i64 8)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP14]], <vscale x 4 x float> [[TMP15]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.fminnm.x4.nxv4f32(<vscale x 4 x float> [[ZDN_COERCE0:%.*]], <vscale x 4 x float> [[ZDN_COERCE1:%.*]], <vscale x 4 x float> [[ZDN_COERCE2:%.*]], <vscale x 4 x float> [[ZDN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svminnm_multi_f32_x4(svfloat32x4_t zdn, svfloat32x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_f32_x4)(zdn, zm);
@ -539,45 +501,35 @@ svfloat32x4_t test_svminnm_multi_f32_x4(svfloat32x4_t zdn, svfloat32x4_t zm) __a
// CHECK-LABEL: @test_svminnm_multi_f64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 2)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP12]], <vscale x 2 x double> [[TMP13]], i64 4)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP14]], <vscale x 2 x double> [[TMP15]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x double> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z25test_svminnm_multi_f64_x413svfloat64x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZDN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.x4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 2)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP12]], <vscale x 2 x double> [[TMP13]], i64 4)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP14]], <vscale x 2 x double> [[TMP15]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x double> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.fminnm.x4.nxv2f64(<vscale x 2 x double> [[ZDN_COERCE0:%.*]], <vscale x 2 x double> [[ZDN_COERCE1:%.*]], <vscale x 2 x double> [[ZDN_COERCE2:%.*]], <vscale x 2 x double> [[ZDN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
svfloat64x4_t test_svminnm_multi_f64_x4(svfloat64x4_t zdn, svfloat64x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svminnm,_f64_x4)(zdn, zm);

144
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_mla.c
View File

@ -19,20 +19,12 @@
// Multi, multi
// CHECK-LABEL: @test_svmla2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmla2_f32j13svfloat32x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32x2_t zm) __arm_streaming __arm_inout("za") {
@ -41,28 +33,12 @@ void test_svmla2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32x2_t zm) __
// CHECK-LABEL: @test_svmla4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmla4_f32j13svfloat32x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32x4_t zm) __arm_streaming __arm_inout("za") {
@ -73,16 +49,12 @@ void test_svmla4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32x4_t zm) __
// Multi, single
// CHECK-LABEL: @test_svmla_single2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmla_single2_f32j13svfloat32x2_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32_t zm) __arm_streaming __arm_inout("za") {
@ -91,20 +63,12 @@ void test_svmla_single2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32_t z
// CHECK-LABEL: @test_svmla_single4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmla_single4_f32j13svfloat32x4_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32_t zm) __arm_streaming __arm_inout("za") {
@ -115,16 +79,12 @@ void test_svmla_single4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32_t z
// Multi, indexed
// CHECK-LABEL: @test_svmla_lane2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmla_lane2_f32j13svfloat32x2_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32_t zm) __arm_streaming __arm_inout("za") {
@ -133,20 +93,12 @@ void test_svmla_lane2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32_t zm)
// CHECK-LABEL: @test_svmla_lane4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmla_lane4_f32j13svfloat32x4_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32_t zm) __arm_streaming __arm_inout("za") {
@ -157,20 +109,12 @@ void test_svmla_lane4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32_t zm)
// Multi, multi
// CHECK-LABEL: @test_svmla2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmla2_f64j13svfloat64x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64x2_t zm) __arm_streaming __arm_inout("za") {
@ -179,28 +123,12 @@ void test_svmla2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64x2_t zm) __
// CHECK-LABEL: @test_svmla4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmla4_f64j13svfloat64x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla4_f64(uint32_t slice_base, svfloat64x4_t zn, svfloat64x4_t zm) __arm_streaming __arm_inout("za") {
@ -211,16 +139,12 @@ void test_svmla4_f64(uint32_t slice_base, svfloat64x4_t zn, svfloat64x4_t zm) __
// Multi, single
// CHECK-LABEL: @test_svmla_single2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmla_single2_f64j13svfloat64x2_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64_t zm) __arm_streaming __arm_inout("za") {
@ -229,20 +153,12 @@ void test_svmla_single2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64_t z
// CHECK-LABEL: @test_svmla_single4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmla_single4_f64j13svfloat64x4_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single4_f64(uint32_t slice_base, svfloat64x4_t zn, svfloat64_t zm) __arm_streaming __arm_inout("za") {
@ -253,16 +169,12 @@ void test_svmla_single4_f64(uint32_t slice_base, svfloat64x4_t zn, svfloat64_t z
// Multi, indexed
// CHECK-LABEL: @test_svmla_lane2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmla_lane2_f64j13svfloat64x2_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64_t zm) __arm_streaming __arm_inout("za") {
@ -271,20 +183,12 @@ void test_svmla_lane2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64_t zm)
// CHECK-LABEL: @test_svmla_lane4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmla_lane4_f64j13svfloat64x4_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.lane.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane4_f64(uint32_t slice_base, svfloat64x4_t zn, svfloat64_t zm) __arm_streaming __arm_inout("za") {

288
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_mlal.c
View File

@ -19,20 +19,12 @@
// Multi, multi
// CHECK-LABEL: @test_svmla2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmla2_f16j13svfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16x2_t zm) __arm_streaming __arm_inout("za")
@ -42,20 +34,12 @@ void test_svmla2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16x2_t zm) __
// CHECK-LABEL: @test_svmla2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z16test_svmla2_bf16j14svbfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16x2_t zm) __arm_streaming __arm_inout("za")
@ -65,20 +49,12 @@ void test_svmla2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16x2_t zm)
// CHECK-LABEL: @test_svmla2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmla2_u16j12svuint16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16x2_t zm) __arm_streaming __arm_inout("za")
@ -88,20 +64,12 @@ void test_svmla2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16x2_t zm) __ar
// CHECK-LABEL: @test_svmla2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmla2_s16j11svint16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla2_s16(uint32_t slice_base, svint16x2_t zn, svint16x2_t zm) __arm_streaming __arm_inout("za")
@ -111,28 +79,12 @@ void test_svmla2_s16(uint32_t slice_base, svint16x2_t zn, svint16x2_t zm) __arm_
// CHECK-LABEL: @test_svmla4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmla4_f16j13svfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16x4_t zm) __arm_streaming __arm_inout("za")
@ -142,28 +94,12 @@ void test_svmla4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16x4_t zm) __
// CHECK-LABEL: @test_svmla4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z16test_svmla4_bf16j14svbfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16x4_t zm) __arm_streaming __arm_inout("za")
@ -173,28 +109,12 @@ void test_svmla4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16x4_t zm)
// CHECK-LABEL: @test_svmla4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE2:%.*]], <vscale x 8 x i16> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmla4_u16j12svuint16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE2:%.*]], <vscale x 8 x i16> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16x4_t zm) __arm_streaming __arm_inout("za")
@ -204,28 +124,12 @@ void test_svmla4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16x4_t zm) __ar
// CHECK-LABEL: @test_svmla4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE2:%.*]], <vscale x 8 x i16> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmla4_s16j11svint16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE2:%.*]], <vscale x 8 x i16> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla4_s16(uint32_t slice_base, svint16x4_t zn, svint16x4_t zm) __arm_streaming __arm_inout("za")
@ -297,16 +201,12 @@ void test_svmla_single1_s16(uint32_t slice_base, svint16_t zn, svint16_t zm) __a
// CHECK-LABEL: @test_svmla_single2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmla_single2_f16j13svfloat16x2_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za")
@ -316,16 +216,12 @@ void test_svmla_single2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16_t z
// CHECK-LABEL: @test_svmla_single2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z23test_svmla_single2_bf16j14svbfloat16x2_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za")
@ -335,16 +231,12 @@ void test_svmla_single2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16_
// CHECK-LABEL: @test_svmla_single2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmla_single2_u16j12svuint16x2_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16_t zm) __arm_streaming __arm_inout("za")
@ -354,16 +246,12 @@ void test_svmla_single2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16_t zm)
// CHECK-LABEL: @test_svmla_single2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmla_single2_s16j11svint16x2_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single2_s16(uint32_t slice_base, svint16x2_t zn, svint16_t zm) __arm_streaming __arm_inout("za")
@ -373,20 +261,12 @@ void test_svmla_single2_s16(uint32_t slice_base, svint16x2_t zn, svint16_t zm) _
// CHECK-LABEL: @test_svmla_single4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmla_single4_f16j13svfloat16x4_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za")
@ -396,20 +276,12 @@ void test_svmla_single4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16_t z
// CHECK-LABEL: @test_svmla_single4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z23test_svmla_single4_bf16j14svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.single.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za")
@ -419,20 +291,12 @@ void test_svmla_single4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16_
// CHECK-LABEL: @test_svmla_single4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmla_single4_u16j12svuint16x4_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16_t zm) __arm_streaming __arm_inout("za")
@ -442,20 +306,12 @@ void test_svmla_single4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16_t zm)
// CHECK-LABEL: @test_svmla_single4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmla_single4_s16j11svint16x4_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmla_single4_s16(uint32_t slice_base, svint16x4_t zn, svint16_t zm) __arm_streaming __arm_inout("za")
@ -529,16 +385,12 @@ void test_svmla_lane1_s16(uint32_t slice_base, svint16_t zn, svint16_t zm) __arm
// CHECK-LABEL: @test_svmla_lane2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmla_lane2_f16j13svfloat16x2_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za")
@ -548,16 +400,12 @@ void test_svmla_lane2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16_t zm)
// CHECK-LABEL: @test_svmla_lane2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z21test_svmla_lane2_bf16j14svbfloat16x2_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za")
@ -567,16 +415,12 @@ void test_svmla_lane2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16_t
// CHECK-LABEL: @test_svmla_lane2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmla_lane2_u16j12svuint16x2_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16_t zm) __arm_streaming __arm_inout("za")
@ -586,16 +430,12 @@ void test_svmla_lane2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16_t zm) _
// CHECK-LABEL: @test_svmla_lane2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmla_lane2_s16j11svint16x2_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane2_s16(uint32_t slice_base, svint16x2_t zn, svint16_t zm) __arm_streaming __arm_inout("za")
@ -605,20 +445,12 @@ void test_svmla_lane2_s16(uint32_t slice_base, svint16x2_t zn, svint16_t zm) __a
// CHECK-LABEL: @test_svmla_lane4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmla_lane4_f16j13svfloat16x4_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za")
@ -628,20 +460,12 @@ void test_svmla_lane4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16_t zm)
// CHECK-LABEL: @test_svmla_lane4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z21test_svmla_lane4_bf16j14svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlal.lane.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za")
@ -651,20 +475,12 @@ void test_svmla_lane4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16_t
// CHECK-LABEL: @test_svmla_lane4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmla_lane4_u16j12svuint16x4_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlal.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16_t zm) __arm_streaming __arm_inout("za")
@ -674,20 +490,12 @@ void test_svmla_lane4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16_t zm) _
// CHECK-LABEL: @test_svmla_lane4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmla_lane4_s16j11svint16x4_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlal.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmla_lane4_s16(uint32_t slice_base, svint16x4_t zn, svint16_t zm) __arm_streaming __arm_inout("za")

720
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_mlall.c
View File

File diff suppressed because it is too large Load Diff

144
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_mls.c
View File

@ -19,20 +19,12 @@
// Multi, multi
// CHECK-LABEL: @test_svmls2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmls2_f32j13svfloat32x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32x2_t zm) __arm_streaming __arm_inout("za") {
@ -41,28 +33,12 @@ void test_svmls2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32x2_t zm) __
// CHECK-LABEL: @test_svmls4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmls4_f32j13svfloat32x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZM]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE2:%.*]], <vscale x 4 x float> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32x4_t zm) __arm_streaming __arm_inout("za") {
@ -73,16 +49,12 @@ void test_svmls4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32x4_t zm) __
// Multi, single
// CHECK-LABEL: @test_svmls_single2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmls_single2_f32j13svfloat32x2_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32_t zm) __arm_streaming __arm_inout("za") {
@ -91,20 +63,12 @@ void test_svmls_single2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32_t z
// CHECK-LABEL: @test_svmls_single4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmls_single4_f32j13svfloat32x4_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32_t zm) __arm_streaming __arm_inout("za") {
@ -115,16 +79,12 @@ void test_svmls_single4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32_t z
// Multi, indexed
// CHECK-LABEL: @test_svmls_lane2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmls_lane2_f32j13svfloat32x2_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32_t zm) __arm_streaming __arm_inout("za") {
@ -133,20 +93,12 @@ void test_svmls_lane2_f32(uint32_t slice_base, svfloat32x2_t zn, svfloat32_t zm)
// CHECK-LABEL: @test_svmls_lane4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmls_lane4_f32j13svfloat32x4_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]], <vscale x 4 x float> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32_t zm) __arm_streaming __arm_inout("za") {
@ -157,20 +109,12 @@ void test_svmls_lane4_f32(uint32_t slice_base, svfloat32x4_t zn, svfloat32_t zm)
// Multi, multi
// CHECK-LABEL: @test_svmls2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmls2_f64j13svfloat64x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64x2_t zm) __arm_streaming __arm_inout("za") {
@ -179,28 +123,12 @@ void test_svmls2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64x2_t zm) __
// CHECK-LABEL: @test_svmls4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmls4_f64j13svfloat64x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZM]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE2:%.*]], <vscale x 2 x double> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls4_f64(uint32_t slice_base, svfloat64x4_t zn, svfloat64x4_t zm) __arm_streaming __arm_inout("za") {
@ -211,16 +139,12 @@ void test_svmls4_f64(uint32_t slice_base, svfloat64x4_t zn, svfloat64x4_t zm) __
// Multi, single
// CHECK-LABEL: @test_svmls_single2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmls_single2_f64j13svfloat64x2_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64_t zm) __arm_streaming __arm_inout("za") {
@ -229,20 +153,12 @@ void test_svmls_single2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64_t z
// CHECK-LABEL: @test_svmls_single4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmls_single4_f64j13svfloat64x4_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single4_f64(uint32_t slice_base, svfloat64x4_t zn, svfloat64_t zm) __arm_streaming __arm_inout("za") {
@ -253,16 +169,12 @@ void test_svmls_single4_f64(uint32_t slice_base, svfloat64x4_t zn, svfloat64_t z
// Multi, indexed
// CHECK-LABEL: @test_svmls_lane2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmls_lane2_f64j13svfloat64x2_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64_t zm) __arm_streaming __arm_inout("za") {
@ -271,20 +183,12 @@ void test_svmls_lane2_f64(uint32_t slice_base, svfloat64x2_t zn, svfloat64_t zm)
// CHECK-LABEL: @test_svmls_lane4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmls_lane4_f64j13svfloat64x4_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.lane.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]], <vscale x 2 x double> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane4_f64(uint32_t slice_base, svfloat64x4_t zn, svfloat64_t zm) __arm_streaming __arm_inout("za") {

288
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_mlsl.c
View File

@ -19,20 +19,12 @@
// Multi, multi
// CHECK-LABEL: @test_svmls2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmls2_f16j13svfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16x2_t zm) __arm_streaming __arm_inout("za")
@ -42,20 +34,12 @@ void test_svmls2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16x2_t zm) __
// CHECK-LABEL: @test_svmls2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z16test_svmls2_bf16j14svbfloat16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16x2_t zm) __arm_streaming __arm_inout("za")
@ -65,20 +49,12 @@ void test_svmls2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16x2_t zm)
// CHECK-LABEL: @test_svmls2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmls2_u16j12svuint16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16x2_t zm) __arm_streaming __arm_inout("za")
@ -88,20 +64,12 @@ void test_svmls2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16x2_t zm) __ar
// CHECK-LABEL: @test_svmls2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmls2_s16j11svint16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls2_s16(uint32_t slice_base, svint16x2_t zn, svint16x2_t zm) __arm_streaming __arm_inout("za")
@ -111,28 +79,12 @@ void test_svmls2_s16(uint32_t slice_base, svint16x2_t zn, svint16x2_t zm) __arm_
// CHECK-LABEL: @test_svmls4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmls4_f16j13svfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE2:%.*]], <vscale x 8 x half> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16x4_t zm) __arm_streaming __arm_inout("za")
@ -142,28 +94,12 @@ void test_svmls4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16x4_t zm) __
// CHECK-LABEL: @test_svmls4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z16test_svmls4_bf16j14svbfloat16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZM]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16x4_t zm) __arm_streaming __arm_inout("za")
@ -173,28 +109,12 @@ void test_svmls4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16x4_t zm)
// CHECK-LABEL: @test_svmls4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE2:%.*]], <vscale x 8 x i16> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmls4_u16j12svuint16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE2:%.*]], <vscale x 8 x i16> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16x4_t zm) __arm_streaming __arm_inout("za")
@ -204,28 +124,12 @@ void test_svmls4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16x4_t zm) __ar
// CHECK-LABEL: @test_svmls4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE2:%.*]], <vscale x 8 x i16> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svmls4_s16j11svint16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE2:%.*]], <vscale x 8 x i16> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls4_s16(uint32_t slice_base, svint16x4_t zn, svint16x4_t zm) __arm_streaming __arm_inout("za")
@ -297,16 +201,12 @@ void test_svmls_single1_s16(uint32_t slice_base, svint16_t zn, svint16_t zm) __a
// CHECK-LABEL: @test_svmls_single2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmls_single2_f16j13svfloat16x2_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za")
@ -316,16 +216,12 @@ void test_svmls_single2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16_t z
// CHECK-LABEL: @test_svmls_single2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z23test_svmls_single2_bf16j14svbfloat16x2_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za")
@ -335,16 +231,12 @@ void test_svmls_single2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16_
// CHECK-LABEL: @test_svmls_single2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmls_single2_u16j12svuint16x2_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16_t zm) __arm_streaming __arm_inout("za")
@ -354,16 +246,12 @@ void test_svmls_single2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16_t zm)
// CHECK-LABEL: @test_svmls_single2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmls_single2_s16j11svint16x2_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.single.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single2_s16(uint32_t slice_base, svint16x2_t zn, svint16_t zm) __arm_streaming __arm_inout("za")
@ -373,20 +261,12 @@ void test_svmls_single2_s16(uint32_t slice_base, svint16x2_t zn, svint16_t zm) _
// CHECK-LABEL: @test_svmls_single4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmls_single4_f16j13svfloat16x4_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za")
@ -396,20 +276,12 @@ void test_svmls_single4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16_t z
// CHECK-LABEL: @test_svmls_single4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z23test_svmls_single4_bf16j14svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.single.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za")
@ -419,20 +291,12 @@ void test_svmls_single4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16_
// CHECK-LABEL: @test_svmls_single4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmls_single4_u16j12svuint16x4_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16_t zm) __arm_streaming __arm_inout("za")
@ -442,20 +306,12 @@ void test_svmls_single4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16_t zm)
// CHECK-LABEL: @test_svmls_single4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z22test_svmls_single4_s16j11svint16x4_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.single.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svmls_single4_s16(uint32_t slice_base, svint16x4_t zn, svint16_t zm) __arm_streaming __arm_inout("za")
@ -529,16 +385,12 @@ void test_svmls_lane1_s16(uint32_t slice_base, svint16_t zn, svint16_t zm) __arm
// CHECK-LABEL: @test_svmls_lane2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmls_lane2_f16j13svfloat16x2_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za")
@ -548,16 +400,12 @@ void test_svmls_lane2_f16(uint32_t slice_base, svfloat16x2_t zn, svfloat16_t zm)
// CHECK-LABEL: @test_svmls_lane2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z21test_svmls_lane2_bf16j14svbfloat16x2_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za")
@ -567,16 +415,12 @@ void test_svmls_lane2_bf16(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16_t
// CHECK-LABEL: @test_svmls_lane2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmls_lane2_u16j12svuint16x2_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16_t zm) __arm_streaming __arm_inout("za")
@ -586,16 +430,12 @@ void test_svmls_lane2_u16(uint32_t slice_base, svuint16x2_t zn, svuint16_t zm) _
// CHECK-LABEL: @test_svmls_lane2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmls_lane2_s16j11svint16x2_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.lane.vg2x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane2_s16(uint32_t slice_base, svint16x2_t zn, svint16_t zm) __arm_streaming __arm_inout("za")
@ -605,20 +445,12 @@ void test_svmls_lane2_s16(uint32_t slice_base, svint16x2_t zn, svint16_t zm) __a
// CHECK-LABEL: @test_svmls_lane4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmls_lane4_f16j13svfloat16x4_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x4.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZN_COERCE2:%.*]], <vscale x 8 x half> [[ZN_COERCE3:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za")
@ -628,20 +460,12 @@ void test_svmls_lane4_f16(uint32_t slice_base, svfloat16x4_t zn, svfloat16_t zm)
// CHECK-LABEL: @test_svmls_lane4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z21test_svmls_lane4_bf16j14svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fmlsl.lane.vg2x4.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za")
@ -651,20 +475,12 @@ void test_svmls_lane4_bf16(uint32_t slice_base, svbfloat16x4_t zn, svbfloat16_t
// CHECK-LABEL: @test_svmls_lane4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmls_lane4_u16j12svuint16x4_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.umlsl.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16_t zm) __arm_streaming __arm_inout("za")
@ -674,20 +490,12 @@ void test_svmls_lane4_u16(uint32_t slice_base, svuint16x4_t zn, svuint16_t zm) _
// CHECK-LABEL: @test_svmls_lane4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svmls_lane4_s16j11svint16x4_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.smlsl.lane.vg2x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 7)
// CPP-CHECK-NEXT: ret void
//
void test_svmls_lane4_s16(uint32_t slice_base, svint16x4_t zn, svint16_t zm) __arm_streaming __arm_inout("za")

721
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_read.c
View File

File diff suppressed because it is too large Load Diff

752
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_sqdmulh.c
View File

@ -19,25 +19,27 @@
// CHECK-LABEL: @test_svqdmulh_single_s8_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv16i8(<vscale x 16 x i8> [[ZDN_COERCE0:%.*]], <vscale x 16 x i8> [[ZDN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z26test_svqdmulh_single_s8_x210svint8x2_tu10__SVInt8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv16i8(<vscale x 16 x i8> [[ZDN_COERCE0:%.*]], <vscale x 16 x i8> [[ZDN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
svint8x2_t test_svqdmulh_single_s8_x2(svint8x2_t zdn, svint8_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_single_s8_x2,,,)(zdn, zm);
@ -45,25 +47,27 @@ svint8x2_t test_svqdmulh_single_s8_x2(svint8x2_t zdn, svint8_t zm) __arm_streami
// CHECK-LABEL: @test_svqdmulh_single_s16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv8i16(<vscale x 8 x i16> [[ZDN_COERCE0:%.*]], <vscale x 8 x i16> [[ZDN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z27test_svqdmulh_single_s16_x211svint16x2_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv8i16(<vscale x 8 x i16> [[ZDN_COERCE0:%.*]], <vscale x 8 x i16> [[ZDN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svint16x2_t test_svqdmulh_single_s16_x2(svint16x2_t zdn, svint16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_single_s16_x2,,,)(zdn, zm);
@ -71,25 +75,27 @@ svint16x2_t test_svqdmulh_single_s16_x2(svint16x2_t zdn, svint16_t zm) __arm_str
// CHECK-LABEL: @test_svqdmulh_single_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv4i32(<vscale x 4 x i32> [[ZDN_COERCE0:%.*]], <vscale x 4 x i32> [[ZDN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z27test_svqdmulh_single_s32_x211svint32x2_tu11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv4i32(<vscale x 4 x i32> [[ZDN_COERCE0:%.*]], <vscale x 4 x i32> [[ZDN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svint32x2_t test_svqdmulh_single_s32_x2(svint32x2_t zdn, svint32_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_single_s32_x2,,,)(zdn, zm);
@ -97,25 +103,27 @@ svint32x2_t test_svqdmulh_single_s32_x2(svint32x2_t zdn, svint32_t zm) __arm_str
// CHECK-LABEL: @test_svqdmulh_single_s64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv2i64(<vscale x 2 x i64> [[ZDN_COERCE0:%.*]], <vscale x 2 x i64> [[ZDN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z27test_svqdmulh_single_s64_x211svint64x2_tu11__SVInt64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.single.vgx2.nxv2i64(<vscale x 2 x i64> [[ZDN_COERCE0:%.*]], <vscale x 2 x i64> [[ZDN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
svint64x2_t test_svqdmulh_single_s64_x2(svint64x2_t zdn, svint64_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_single_s64_x2,,,)(zdn, zm);
@ -125,37 +133,35 @@ svint64x2_t test_svqdmulh_single_s64_x2(svint64x2_t zdn, svint64_t zm) __arm_str
// CHECK-LABEL: @test_svqdmulh_single_s8_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 48)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]], i64 32)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv16i8(<vscale x 16 x i8> [[ZDN_COERCE0:%.*]], <vscale x 16 x i8> [[ZDN_COERCE1:%.*]], <vscale x 16 x i8> [[ZDN_COERCE2:%.*]], <vscale x 16 x i8> [[ZDN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z26test_svqdmulh_single_s8_x410svint8x4_tu10__SVInt8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 48)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]], i64 32)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv16i8(<vscale x 16 x i8> [[ZDN_COERCE0:%.*]], <vscale x 16 x i8> [[ZDN_COERCE1:%.*]], <vscale x 16 x i8> [[ZDN_COERCE2:%.*]], <vscale x 16 x i8> [[ZDN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CPP-CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
svint8x4_t test_svqdmulh_single_s8_x4(svint8x4_t zdn, svint8_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_single_s8_x4,,,)(zdn, zm);
@ -163,37 +169,35 @@ svint8x4_t test_svqdmulh_single_s8_x4(svint8x4_t zdn, svint8_t zm) __arm_streami
// CHECK-LABEL: @test_svqdmulh_single_s16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 16)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv8i16(<vscale x 8 x i16> [[ZDN_COERCE0:%.*]], <vscale x 8 x i16> [[ZDN_COERCE1:%.*]], <vscale x 8 x i16> [[ZDN_COERCE2:%.*]], <vscale x 8 x i16> [[ZDN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z27test_svqdmulh_single_s16_x411svint16x4_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 16)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv8i16(<vscale x 8 x i16> [[ZDN_COERCE0:%.*]], <vscale x 8 x i16> [[ZDN_COERCE1:%.*]], <vscale x 8 x i16> [[ZDN_COERCE2:%.*]], <vscale x 8 x i16> [[ZDN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svint16x4_t test_svqdmulh_single_s16_x4(svint16x4_t zdn, svint16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_single_s16_x4,,,)(zdn, zm);
@ -201,37 +205,35 @@ svint16x4_t test_svqdmulh_single_s16_x4(svint16x4_t zdn, svint16_t zm) __arm_str
// CHECK-LABEL: @test_svqdmulh_single_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv4i32(<vscale x 4 x i32> [[ZDN_COERCE0:%.*]], <vscale x 4 x i32> [[ZDN_COERCE1:%.*]], <vscale x 4 x i32> [[ZDN_COERCE2:%.*]], <vscale x 4 x i32> [[ZDN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z27test_svqdmulh_single_s32_x411svint32x4_tu11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv4i32(<vscale x 4 x i32> [[ZDN_COERCE0:%.*]], <vscale x 4 x i32> [[ZDN_COERCE1:%.*]], <vscale x 4 x i32> [[ZDN_COERCE2:%.*]], <vscale x 4 x i32> [[ZDN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svint32x4_t test_svqdmulh_single_s32_x4(svint32x4_t zdn, svint32_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_single_s32_x4,,,)(zdn, zm);
@ -239,37 +241,35 @@ svint32x4_t test_svqdmulh_single_s32_x4(svint32x4_t zdn, svint32_t zm) __arm_str
// CHECK-LABEL: @test_svqdmulh_single_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 4)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv2i64(<vscale x 2 x i64> [[ZDN_COERCE0:%.*]], <vscale x 2 x i64> [[ZDN_COERCE1:%.*]], <vscale x 2 x i64> [[ZDN_COERCE2:%.*]], <vscale x 2 x i64> [[ZDN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z27test_svqdmulh_single_s64_x411svint64x4_tu11__SVInt64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 4)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.single.vgx4.nxv2i64(<vscale x 2 x i64> [[ZDN_COERCE0:%.*]], <vscale x 2 x i64> [[ZDN_COERCE1:%.*]], <vscale x 2 x i64> [[ZDN_COERCE2:%.*]], <vscale x 2 x i64> [[ZDN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
svint64x4_t test_svqdmulh_single_s64_x4(svint64x4_t zdn, svint64_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_single_s64_x4,,,)(zdn, zm);
@ -279,29 +279,27 @@ svint64x4_t test_svqdmulh_single_s64_x4(svint64x4_t zdn, svint64_t zm) __arm_str
// CHECK-LABEL: @test_svqdmulh_s8_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv16i8(<vscale x 16 x i8> [[ZDN_COERCE0:%.*]], <vscale x 16 x i8> [[ZDN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM_COERCE0:%.*]], <vscale x 16 x i8> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z19test_svqdmulh_s8_x210svint8x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv16i8(<vscale x 16 x i8> [[ZDN_COERCE0:%.*]], <vscale x 16 x i8> [[ZDN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM_COERCE0:%.*]], <vscale x 16 x i8> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
svint8x2_t test_svqdmulh_s8_x2(svint8x2_t zdn, svint8x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_s8_x2,,,)(zdn, zm);
@ -309,29 +307,27 @@ svint8x2_t test_svqdmulh_s8_x2(svint8x2_t zdn, svint8x2_t zm) __arm_streaming {
// CHECK-LABEL: @test_svqdmulh_s16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv8i16(<vscale x 8 x i16> [[ZDN_COERCE0:%.*]], <vscale x 8 x i16> [[ZDN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z20test_svqdmulh_s16_x211svint16x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv8i16(<vscale x 8 x i16> [[ZDN_COERCE0:%.*]], <vscale x 8 x i16> [[ZDN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svint16x2_t test_svqdmulh_s16_x2(svint16x2_t zdn, svint16x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_s16_x2,,,)(zdn, zm);
@ -339,29 +335,27 @@ svint16x2_t test_svqdmulh_s16_x2(svint16x2_t zdn, svint16x2_t zm) __arm_streamin
// CHECK-LABEL: @test_svqdmulh_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv4i32(<vscale x 4 x i32> [[ZDN_COERCE0:%.*]], <vscale x 4 x i32> [[ZDN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z20test_svqdmulh_s32_x211svint32x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv4i32(<vscale x 4 x i32> [[ZDN_COERCE0:%.*]], <vscale x 4 x i32> [[ZDN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svint32x2_t test_svqdmulh_s32_x2(svint32x2_t zdn, svint32x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_s32_x2,,,)(zdn, zm);
@ -369,29 +363,27 @@ svint32x2_t test_svqdmulh_s32_x2(svint32x2_t zdn, svint32x2_t zm) __arm_streamin
// CHECK-LABEL: @test_svqdmulh_s64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv2i64(<vscale x 2 x i64> [[ZDN_COERCE0:%.*]], <vscale x 2 x i64> [[ZDN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z20test_svqdmulh_s64_x211svint64x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.vgx2.nxv2i64(<vscale x 2 x i64> [[ZDN_COERCE0:%.*]], <vscale x 2 x i64> [[ZDN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
svint64x2_t test_svqdmulh_s64_x2(svint64x2_t zdn, svint64x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_s64_x2,,,)(zdn, zm);
@ -401,45 +393,35 @@ svint64x2_t test_svqdmulh_s64_x2(svint64x2_t zdn, svint64x2_t zm) __arm_streamin
// CHECK-LABEL: @test_svqdmulh_s8_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 48)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZM]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZM]], i64 48)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 16)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP13]], i64 32)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv16i8(<vscale x 16 x i8> [[ZDN_COERCE0:%.*]], <vscale x 16 x i8> [[ZDN_COERCE1:%.*]], <vscale x 16 x i8> [[ZDN_COERCE2:%.*]], <vscale x 16 x i8> [[ZDN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM_COERCE0:%.*]], <vscale x 16 x i8> [[ZM_COERCE1:%.*]], <vscale x 16 x i8> [[ZM_COERCE2:%.*]], <vscale x 16 x i8> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z19test_svqdmulh_s8_x410svint8x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZDN]], i64 48)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZM]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZM]], i64 48)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 16)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP13]], i64 32)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv16i8(<vscale x 16 x i8> [[ZDN_COERCE0:%.*]], <vscale x 16 x i8> [[ZDN_COERCE1:%.*]], <vscale x 16 x i8> [[ZDN_COERCE2:%.*]], <vscale x 16 x i8> [[ZDN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM_COERCE0:%.*]], <vscale x 16 x i8> [[ZM_COERCE1:%.*]], <vscale x 16 x i8> [[ZM_COERCE2:%.*]], <vscale x 16 x i8> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CPP-CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
svint8x4_t test_svqdmulh_s8_x4(svint8x4_t zdn, svint8x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_s8_x4,,,)(zdn, zm);
@ -447,45 +429,35 @@ svint8x4_t test_svqdmulh_s8_x4(svint8x4_t zdn, svint8x4_t zm) __arm_streaming {
// CHECK-LABEL: @test_svqdmulh_s16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 24)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 8)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP12]], <vscale x 8 x i16> [[TMP13]], i64 16)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP14]], <vscale x 8 x i16> [[TMP15]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv8i16(<vscale x 8 x i16> [[ZDN_COERCE0:%.*]], <vscale x 8 x i16> [[ZDN_COERCE1:%.*]], <vscale x 8 x i16> [[ZDN_COERCE2:%.*]], <vscale x 8 x i16> [[ZDN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE2:%.*]], <vscale x 8 x i16> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z20test_svqdmulh_s16_x411svint16x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZDN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZM]], i64 24)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 8)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP12]], <vscale x 8 x i16> [[TMP13]], i64 16)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP14]], <vscale x 8 x i16> [[TMP15]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv8i16(<vscale x 8 x i16> [[ZDN_COERCE0:%.*]], <vscale x 8 x i16> [[ZDN_COERCE1:%.*]], <vscale x 8 x i16> [[ZDN_COERCE2:%.*]], <vscale x 8 x i16> [[ZDN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE2:%.*]], <vscale x 8 x i16> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svint16x4_t test_svqdmulh_s16_x4(svint16x4_t zdn, svint16x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_s16_x4,,,)(zdn, zm);
@ -493,45 +465,35 @@ svint16x4_t test_svqdmulh_s16_x4(svint16x4_t zdn, svint16x4_t zm) __arm_streamin
// CHECK-LABEL: @test_svqdmulh_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 12)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 4)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP12]], <vscale x 4 x i32> [[TMP13]], i64 8)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP14]], <vscale x 4 x i32> [[TMP15]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv4i32(<vscale x 4 x i32> [[ZDN_COERCE0:%.*]], <vscale x 4 x i32> [[ZDN_COERCE1:%.*]], <vscale x 4 x i32> [[ZDN_COERCE2:%.*]], <vscale x 4 x i32> [[ZDN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE2:%.*]], <vscale x 4 x i32> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z20test_svqdmulh_s32_x411svint32x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZDN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 12)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 4)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP12]], <vscale x 4 x i32> [[TMP13]], i64 8)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP14]], <vscale x 4 x i32> [[TMP15]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv4i32(<vscale x 4 x i32> [[ZDN_COERCE0:%.*]], <vscale x 4 x i32> [[ZDN_COERCE1:%.*]], <vscale x 4 x i32> [[ZDN_COERCE2:%.*]], <vscale x 4 x i32> [[ZDN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE2:%.*]], <vscale x 4 x i32> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svint32x4_t test_svqdmulh_s32_x4(svint32x4_t zdn, svint32x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_s32_x4,,,)(zdn, zm);
@ -539,45 +501,35 @@ svint32x4_t test_svqdmulh_s32_x4(svint32x4_t zdn, svint32x4_t zm) __arm_streamin
// CHECK-LABEL: @test_svqdmulh_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 6)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 2)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP12]], <vscale x 2 x i64> [[TMP13]], i64 4)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP14]], <vscale x 2 x i64> [[TMP15]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv2i64(<vscale x 2 x i64> [[ZDN_COERCE0:%.*]], <vscale x 2 x i64> [[ZDN_COERCE1:%.*]], <vscale x 2 x i64> [[ZDN_COERCE2:%.*]], <vscale x 2 x i64> [[ZDN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE2:%.*]], <vscale x 2 x i64> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z20test_svqdmulh_s64_x411svint64x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZDN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 6)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 2)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP12]], <vscale x 2 x i64> [[TMP13]], i64 4)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP14]], <vscale x 2 x i64> [[TMP15]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sqdmulh.vgx4.nxv2i64(<vscale x 2 x i64> [[ZDN_COERCE0:%.*]], <vscale x 2 x i64> [[ZDN_COERCE1:%.*]], <vscale x 2 x i64> [[ZDN_COERCE2:%.*]], <vscale x 2 x i64> [[ZDN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE2:%.*]], <vscale x 2 x i64> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
svint64x4_t test_svqdmulh_s64_x4(svint64x4_t zdn, svint64x4_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svqdmulh,_s64_x4,,,)(zdn, zm);

328
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_sub.c
View File

@ -24,16 +24,12 @@
// x2
// CHECK-LABEL: @test_svsub_write_single2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svsub_write_single2_s32j11svint32x2_tu11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_single2_s32(uint32_t slice_base, svint32x2_t zn, svint32_t zm) __arm_streaming __arm_inout("za") {
@ -42,16 +38,12 @@ void test_svsub_write_single2_s32(uint32_t slice_base, svint32x2_t zn, svint32_t
// CHECK-LABEL: @test_svsub_write_single2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svsub_write_single2_u32j12svuint32x2_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_single2_u32(uint32_t slice_base, svuint32x2_t zn, svuint32_t zm) __arm_streaming __arm_inout("za") {
@ -60,16 +52,12 @@ void test_svsub_write_single2_u32(uint32_t slice_base, svuint32x2_t zn, svuint32
// CHECK-LABEL: @test_svsub_write_single2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svsub_write_single2_s64j11svint64x2_tu11__SVInt64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_single2_s64(uint32_t slice_base, svint64x2_t zn, svint64_t zm) __arm_streaming __arm_inout("za") {
@ -78,16 +66,12 @@ void test_svsub_write_single2_s64(uint32_t slice_base, svint64x2_t zn, svint64_t
// CHECK-LABEL: @test_svsub_write_single2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svsub_write_single2_u64j12svuint64x2_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_single2_u64(uint32_t slice_base, svuint64x2_t zn, svuint64_t zm) __arm_streaming __arm_inout("za") {
@ -98,20 +82,12 @@ void test_svsub_write_single2_u64(uint32_t slice_base, svuint64x2_t zn, svuint64
// CHECK-LABEL: @test_svsub_write_single4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svsub_write_single4_s32j11svint32x4_tu11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_single4_s32(uint32_t slice_base, svint32x4_t zn, svint32_t zm) __arm_streaming __arm_inout("za") {
@ -120,20 +96,12 @@ void test_svsub_write_single4_s32(uint32_t slice_base, svint32x4_t zn, svint32_t
// CHECK-LABEL: @test_svsub_write_single4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svsub_write_single4_u32j12svuint32x4_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_single4_u32(uint32_t slice_base, svuint32x4_t zn, svuint32_t zm) __arm_streaming __arm_inout("za") {
@ -142,20 +110,12 @@ void test_svsub_write_single4_u32(uint32_t slice_base, svuint32x4_t zn, svuint32
// CHECK-LABEL: @test_svsub_write_single4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svsub_write_single4_s64j11svint64x4_tu11__SVInt64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_single4_s64(uint32_t slice_base, svint64x4_t zn, svint64_t zm) __arm_streaming __arm_inout("za") {
@ -164,20 +124,12 @@ void test_svsub_write_single4_s64(uint32_t slice_base, svint64x4_t zn, svint64_t
// CHECK-LABEL: @test_svsub_write_single4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z28test_svsub_write_single4_u64j12svuint64x4_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.single.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_single4_u64(uint32_t slice_base, svuint64x4_t zn, svuint64_t zm) __arm_streaming __arm_inout("za") {
@ -192,20 +144,12 @@ void test_svsub_write_single4_u64(uint32_t slice_base, svuint64x4_t zn, svuint64
// CHECK-LABEL: @test_svsub_write_multi2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svsub_write_multi2_s32j11svint32x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_multi2_s32(uint32_t slice_base, svint32x2_t zn, svint32x2_t zm) __arm_streaming __arm_inout("za") {
@ -214,20 +158,12 @@ void test_svsub_write_multi2_s32(uint32_t slice_base, svint32x2_t zn, svint32x2_
// CHECK-LABEL: @test_svsub_write_multi2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svsub_write_multi2_u32j12svuint32x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_multi2_u32(uint32_t slice_base, svuint32x2_t zn, svuint32x2_t zm) __arm_streaming __arm_inout("za") {
@ -236,20 +172,12 @@ void test_svsub_write_multi2_u32(uint32_t slice_base, svuint32x2_t zn, svuint32x
// CHECK-LABEL: @test_svsub_write_multi2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svsub_write_multi2_s64j11svint64x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_multi2_s64(uint32_t slice_base, svint64x2_t zn, svint64x2_t zm) __arm_streaming __arm_inout("za") {
@ -258,20 +186,12 @@ void test_svsub_write_multi2_s64(uint32_t slice_base, svint64x2_t zn, svint64x2_
// CHECK-LABEL: @test_svsub_write_multi2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svsub_write_multi2_u64j12svuint64x2_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_multi2_u64(uint32_t slice_base, svuint64x2_t zn, svuint64x2_t zm) __arm_streaming __arm_inout("za") {
@ -282,28 +202,12 @@ void test_svsub_write_multi2_u64(uint32_t slice_base, svuint64x2_t zn, svuint64x
// CHECK-LABEL: @test_svsub_write_multi4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE2:%.*]], <vscale x 4 x i32> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svsub_write_multi4_s32j11svint32x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE2:%.*]], <vscale x 4 x i32> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_multi4_s32(uint32_t slice_base, svint32x4_t zn, svint32x4_t zm) __arm_streaming __arm_inout("za") {
@ -312,28 +216,12 @@ void test_svsub_write_multi4_s32(uint32_t slice_base, svint32x4_t zn, svint32x4_
// CHECK-LABEL: @test_svsub_write_multi4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE2:%.*]], <vscale x 4 x i32> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svsub_write_multi4_u32j12svuint32x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZM]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE2:%.*]], <vscale x 4 x i32> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_multi4_u32(uint32_t slice_base, svuint32x4_t zn, svuint32x4_t zm) __arm_streaming __arm_inout("za") {
@ -342,28 +230,12 @@ void test_svsub_write_multi4_u32(uint32_t slice_base, svuint32x4_t zn, svuint32x
// CHECK-LABEL: @test_svsub_write_multi4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE2:%.*]], <vscale x 2 x i64> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svsub_write_multi4_s64j11svint64x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE2:%.*]], <vscale x 2 x i64> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_multi4_s64(uint32_t slice_base, svint64x4_t zn, svint64x4_t zm) __arm_streaming __arm_inout("za") {
@ -372,28 +244,12 @@ void test_svsub_write_multi4_s64(uint32_t slice_base, svint64x4_t zn, svint64x4_
// CHECK-LABEL: @test_svsub_write_multi4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE2:%.*]], <vscale x 2 x i64> [[ZM_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z27test_svsub_write_multi4_u64j12svuint64x4_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZM]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.write.za.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE2:%.*]], <vscale x 2 x i64> [[ZM_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_write_multi4_u64(uint32_t slice_base, svuint64x4_t zn, svuint64x4_t zm) __arm_streaming __arm_inout("za") {
@ -408,16 +264,12 @@ void test_svsub_write_multi4_u64(uint32_t slice_base, svuint64x4_t zn, svuint64x
// CHECK-LABEL: @test_svsub_za32_vg1x2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za32_vg1x2_f32j13svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za32_vg1x2_f32(uint32_t slice_base, svfloat32x2_t zn) __arm_streaming __arm_inout("za") {
@ -426,16 +278,12 @@ void test_svsub_za32_vg1x2_f32(uint32_t slice_base, svfloat32x2_t zn) __arm_stre
// CHECK-LABEL: @test_svsub_za32_vg1x2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za32_vg1x2_s32j11svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za32_vg1x2_s32(uint32_t slice_base, svint32x2_t zn) __arm_streaming __arm_inout("za") {
@ -444,16 +292,12 @@ void test_svsub_za32_vg1x2_s32(uint32_t slice_base, svint32x2_t zn) __arm_stream
// CHECK-LABEL: @test_svsub_za32_vg1x2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za32_vg1x2_u32j12svuint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x2.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za32_vg1x2_u32(uint32_t slice_base, svuint32x2_t zn) __arm_streaming __arm_inout("za") {
@ -462,16 +306,12 @@ void test_svsub_za32_vg1x2_u32(uint32_t slice_base, svuint32x2_t zn) __arm_strea
// CHECK-LABEL: @test_svsub_za64_vg1x2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za64_vg1x2_f64j13svfloat64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za64_vg1x2_f64(uint32_t slice_base, svfloat64x2_t zn) __arm_streaming __arm_inout("za") {
@ -480,16 +320,12 @@ void test_svsub_za64_vg1x2_f64(uint32_t slice_base, svfloat64x2_t zn) __arm_stre
// CHECK-LABEL: @test_svsub_za64_vg1x2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za64_vg1x2_s64j11svint64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za64_vg1x2_s64(uint32_t slice_base, svint64x2_t zn) __arm_streaming __arm_inout("za") {
@ -498,16 +334,12 @@ void test_svsub_za64_vg1x2_s64(uint32_t slice_base, svint64x2_t zn) __arm_stream
// CHECK-LABEL: @test_svsub_za64_vg1x2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za64_vg1x2_u64j12svuint64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x2.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za64_vg1x2_u64(uint32_t slice_base, svuint64x2_t zn) __arm_streaming __arm_inout("za") {
@ -518,20 +350,12 @@ void test_svsub_za64_vg1x2_u64(uint32_t slice_base, svuint64x2_t zn) __arm_strea
// CHECK-LABEL: @test_svsub_za32_vg1x4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za32_vg1x4_f32j13svfloat32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4f32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZN_COERCE2:%.*]], <vscale x 4 x float> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za32_vg1x4_f32(uint32_t slice_base, svfloat32x4_t zn) __arm_streaming __arm_inout("za") {
@ -540,20 +364,12 @@ void test_svsub_za32_vg1x4_f32(uint32_t slice_base, svfloat32x4_t zn) __arm_stre
// CHECK-LABEL: @test_svsub_za32_vg1x4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za32_vg1x4_s32j11svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za32_vg1x4_s32(uint32_t slice_base, svint32x4_t zn) __arm_streaming __arm_inout("za") {
@ -562,20 +378,12 @@ void test_svsub_za32_vg1x4_s32(uint32_t slice_base, svint32x4_t zn) __arm_stream
// CHECK-LABEL: @test_svsub_za32_vg1x4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za32_vg1x4_u32j12svuint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za32.vg1x4.nxv4i32(i32 [[SLICE_BASE:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za32_vg1x4_u32(uint32_t slice_base, svuint32x4_t zn) __arm_streaming __arm_inout("za") {
@ -584,20 +392,12 @@ void test_svsub_za32_vg1x4_u32(uint32_t slice_base, svuint32x4_t zn) __arm_strea
// CHECK-LABEL: @test_svsub_za64_vg1x4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za64_vg1x4_f64j13svfloat64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2f64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZN_COERCE2:%.*]], <vscale x 2 x double> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za64_vg1x4_f64(uint32_t slice_base, svfloat64x4_t zn) __arm_streaming __arm_inout("za") {
@ -606,20 +406,12 @@ void test_svsub_za64_vg1x4_f64(uint32_t slice_base, svfloat64x4_t zn) __arm_stre
// CHECK-LABEL: @test_svsub_za64_vg1x4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za64_vg1x4_s64j11svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za64_vg1x4_s64(uint32_t slice_base, svint64x4_t zn) __arm_streaming __arm_inout("za") {
@ -628,20 +420,12 @@ void test_svsub_za64_vg1x4_s64(uint32_t slice_base, svint64x4_t zn) __arm_stream
// CHECK-LABEL: @test_svsub_za64_vg1x4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z25test_svsub_za64_vg1x4_u64j12svuint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.sub.za64.vg1x4.nxv2i64(i32 [[SLICE_BASE:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svsub_za64_vg1x4_u64(uint32_t slice_base, svuint64x4_t zn) __arm_streaming __arm_inout("za") {

60
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_unpkx2.c
View File

@ -19,21 +19,27 @@
// CHECK-LABEL: @test_svunpk_s16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sunpk.x2.nxv8i16(<vscale x 16 x i8> [[ZN:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_s16_x2u10__SVInt8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sunpk.x2.nxv8i16(<vscale x 16 x i8> [[ZN:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svint16x2_t test_svunpk_s16_x2(svint8_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_s16,_s8_x2)(zn);
@ -41,21 +47,27 @@ svint16x2_t test_svunpk_s16_x2(svint8_t zn) __arm_streaming {
// CHECK-LABEL: @test_svunpk_u16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.uunpk.x2.nxv8i16(<vscale x 16 x i8> [[ZN:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_u16_x2u11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.uunpk.x2.nxv8i16(<vscale x 16 x i8> [[ZN:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svuint16x2_t test_svunpk_u16_x2(svuint8_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_u16,_u8_x2)(zn);
@ -63,21 +75,27 @@ svuint16x2_t test_svunpk_u16_x2(svuint8_t zn) __arm_streaming {
// CHECK-LABEL: @test_svunpk_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sunpk.x2.nxv4i32(<vscale x 8 x i16> [[ZN:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_s32_x2u11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sunpk.x2.nxv4i32(<vscale x 8 x i16> [[ZN:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svint32x2_t test_svunpk_s32_x2(svint16_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_s32,_s16_x2)(zn);
@ -85,21 +103,27 @@ svint32x2_t test_svunpk_s32_x2(svint16_t zn) __arm_streaming {
// CHECK-LABEL: @test_svunpk_u32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.uunpk.x2.nxv4i32(<vscale x 8 x i16> [[ZN:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_u32_x2u12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.uunpk.x2.nxv4i32(<vscale x 8 x i16> [[ZN:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svuint32x2_t test_svunpk_u32_x2(svuint16_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_u32,_u16_x2)(zn);
@ -107,21 +131,27 @@ svuint32x2_t test_svunpk_u32_x2(svuint16_t zn) __arm_streaming {
// CHECK-LABEL: @test_svunpk_s64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sunpk.x2.nxv2i64(<vscale x 4 x i32> [[ZN:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP4]]
// CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_s64_x2u11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sunpk.x2.nxv2i64(<vscale x 4 x i32> [[ZN:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
svint64x2_t test_svunpk_s64_x2(svint32_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_s64,_s32_x2)(zn);
@ -129,21 +159,27 @@ svint64x2_t test_svunpk_s64_x2(svint32_t zn) __arm_streaming {
// CHECK-LABEL: @test_svunpk_u64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.uunpk.x2.nxv2i64(<vscale x 4 x i32> [[ZN:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP4]]
// CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_u64_x2u12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.uunpk.x2.nxv2i64(<vscale x 4 x i32> [[ZN:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
svuint64x2_t test_svunpk_u64_x2(svuint32_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_u64,_u32_x2)(zn);

300
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_unpkx4.c
View File

@ -19,33 +19,35 @@
// CHECK-LABEL: @test_svunpk_s16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sunpk.x4.nxv8i16(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP10]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sunpk.x4.nxv8i16(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_s16_x410svint8x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sunpk.x4.nxv8i16(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP10]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sunpk.x4.nxv8i16(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svint16x4_t test_svunpk_s16_x4(svint8x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_s16,_s8_x4)(zn);
@ -53,33 +55,35 @@ svint16x4_t test_svunpk_s16_x4(svint8x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svunpk_u16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.uunpk.x4.nxv8i16(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP10]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.uunpk.x4.nxv8i16(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_u16_x411svuint8x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.uunpk.x4.nxv8i16(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP10]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.uunpk.x4.nxv8i16(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svuint16x4_t test_svunpk_u16_x4(svuint8x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_u16,_u8_x4)(zn);
@ -87,33 +91,35 @@ svuint16x4_t test_svunpk_u16_x4(svuint8x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svunpk_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sunpk.x4.nxv4i32(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP10]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sunpk.x4.nxv4i32(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_s32_x411svint16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sunpk.x4.nxv4i32(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP10]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sunpk.x4.nxv4i32(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svint32x4_t test_svunpk_s32_x4(svint16x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_s32,_s16_x4)(zn);
@ -121,33 +127,35 @@ svint32x4_t test_svunpk_s32_x4(svint16x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svunpk_u32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.uunpk.x4.nxv4i32(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP10]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.uunpk.x4.nxv4i32(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_u32_x412svuint16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.uunpk.x4.nxv4i32(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP10]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.uunpk.x4.nxv4i32(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svuint32x4_t test_svunpk_u32_x4(svuint16x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_u32,_u16_x4)(zn);
@ -155,33 +163,35 @@ svuint32x4_t test_svunpk_u32_x4(svuint16x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svunpk_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sunpk.x4.nxv2i64(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP10]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sunpk.x4.nxv2i64(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_s64_x411svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sunpk.x4.nxv2i64(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP10]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sunpk.x4.nxv2i64(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
svint64x4_t test_svunpk_s64_x4(svint32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_s64,_s32_x4)(zn);
@ -189,33 +199,35 @@ svint64x4_t test_svunpk_s64_x4(svint32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svunpk_u64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.uunpk.x4.nxv2i64(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP10]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.uunpk.x4.nxv2i64(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svunpk_u64_x412svuint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.uunpk.x4.nxv2i64(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP10]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.uunpk.x4.nxv2i64(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
svuint64x4_t test_svunpk_u64_x4(svuint32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svunpk_u64,_u32_x4)(zn);

104
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_vdot.c
View File

@ -16,16 +16,12 @@
// CHECK-LABEL: @test_svvdot_lane_za32_bf16_vg1x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z32test_svvdot_lane_za32_bf16_vg1x2j14svbfloat16x2_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8bf16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svvdot_lane_za32_bf16_vg1x2(uint32_t slice_base, svbfloat16x2_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -34,16 +30,12 @@ void test_svvdot_lane_za32_bf16_vg1x2(uint32_t slice_base, svbfloat16x2_t zn, sv
// CHECK-LABEL: @test_svvdot_lane_za32_f16_vg1x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z31test_svvdot_lane_za32_f16_vg1x2j13svfloat16x2_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8f16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svvdot_lane_za32_f16_vg1x2(uint32_t slice_base, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
@ -52,16 +44,12 @@ void test_svvdot_lane_za32_f16_vg1x2(uint32_t slice_base, svfloat16x2_t zn, svfl
// CHECK-LABEL: @test_svvdot_lane_za32_s16_vg1x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z31test_svvdot_lane_za32_s16_vg1x2j11svint16x2_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svvdot_lane_za32_s16_vg1x2(uint32_t slice_base, svint16x2_t zn, svint16_t zm) __arm_streaming __arm_inout("za") {
@ -70,16 +58,12 @@ void test_svvdot_lane_za32_s16_vg1x2(uint32_t slice_base, svint16x2_t zn, svint1
// CHECK-LABEL: @test_svvdot_lane_za32_u16_vg1x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z31test_svvdot_lane_za32_u16_vg1x2j12svuint16x2_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x2.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svvdot_lane_za32_u16_vg1x2(uint32_t slice_base, svuint16x2_t zn, svuint16_t zm) __arm_streaming __arm_inout("za") {
@ -88,20 +72,12 @@ void test_svvdot_lane_za32_u16_vg1x2(uint32_t slice_base, svuint16x2_t zn, svuin
// CHECK-LABEL: @test_svvdot_lane_za32_s8_vg1x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z30test_svvdot_lane_za32_s8_vg1x4j10svint8x4_tu10__SVInt8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svvdot_lane_za32_s8_vg1x4(uint32_t slice_base, svint8x4_t zn, svint8_t zm) __arm_streaming __arm_inout("za") {
@ -110,20 +86,12 @@ void test_svvdot_lane_za32_s8_vg1x4(uint32_t slice_base, svint8x4_t zn, svint8_t
// CHECK-LABEL: @test_svvdot_lane_za32_u8_vg1x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z30test_svvdot_lane_za32_u8_vg1x4j11svuint8x4_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svvdot_lane_za32_u8_vg1x4(uint32_t slice_base, svuint8x4_t zn, svuint8_t zm) __arm_streaming __arm_inout("za") {
@ -132,20 +100,12 @@ void test_svvdot_lane_za32_u8_vg1x4(uint32_t slice_base, svuint8x4_t zn, svuint8
// CHECK-LABEL: @test_svvdot_lane_za64_s16_vg1x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za64.vg1x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 1)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za64.vg1x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 1)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z31test_svvdot_lane_za64_s16_vg1x4j11svint16x4_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za64.vg1x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.svdot.lane.za64.vg1x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: ret void
//
void test_svvdot_lane_za64_s16_vg1x4(uint32_t slice_base, svint16x4_t zn, svint16_t zm) __arm_streaming __arm_inout("za") {
@ -154,20 +114,12 @@ void test_svvdot_lane_za64_s16_vg1x4(uint32_t slice_base, svint16x4_t zn, svint1
// CHECK-LABEL: @test_svvdot_lane_za64_u16_vg1x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za64.vg1x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 1)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za64.vg1x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 1)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z31test_svvdot_lane_za64_u16_vg1x4j12svuint16x4_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za64.vg1x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.uvdot.lane.za64.vg1x4.nxv8i16(i32 [[SLICE_BASE:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]], i32 1)
// CPP-CHECK-NEXT: ret void
//
void test_svvdot_lane_za64_u16_vg1x4(uint32_t slice_base, svuint16x4_t zn, svuint16_t zm) __arm_streaming __arm_inout("za") {
@ -177,20 +129,12 @@ void test_svvdot_lane_za64_u16_vg1x4(uint32_t slice_base, svuint16x4_t zn, svuin
// CHECK-LABEL: @test_svsuvdot_lane_za32_s8_vg1x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.suvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.suvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z32test_svsuvdot_lane_za32_s8_vg1x4j10svint8x4_tu10__SVInt8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.suvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.suvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svsuvdot_lane_za32_s8_vg1x4(uint32_t slice_base, svint8x4_t zn, svint8_t zm) __arm_streaming __arm_inout("za") {
@ -200,20 +144,12 @@ void test_svsuvdot_lane_za32_s8_vg1x4(uint32_t slice_base, svint8x4_t zn, svint8
// CHECK-LABEL: @test_svusvdot_lane_za32_u8_vg1x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.usvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CHECK-NEXT: tail call void @llvm.aarch64.sme.usvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z32test_svusvdot_lane_za32_u8_vg1x4j11svuint8x4_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.usvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sme.usvdot.lane.za32.vg1x4.nxv16i8(i32 [[SLICE_BASE:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]], i32 3)
// CPP-CHECK-NEXT: ret void
//
void test_svusvdot_lane_za32_u8_vg1x4(uint32_t slice_base, svuint8x4_t zn, svuint8_t zm) __arm_streaming __arm_inout("za") {

704
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_vector_add.c
View File

@ -25,25 +25,27 @@
// CHECK-LABEL: @test_svadd_vector_single2_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x2.nxv16i8(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z28test_svadd_vector_single2_s810svint8x2_tu10__SVInt8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x2.nxv16i8(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
svint8x2_t test_svadd_vector_single2_s8(svint8x2_t zn, svint8_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_s8_x2,,,)(zn, zm);
@ -51,25 +53,27 @@ svint8x2_t test_svadd_vector_single2_s8(svint8x2_t zn, svint8_t zm) __arm_stream
// CHECK-LABEL: @test_svadd_vector_single2_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x2.nxv16i8(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z28test_svadd_vector_single2_u811svuint8x2_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x2.nxv16i8(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
svuint8x2_t test_svadd_vector_single2_u8(svuint8x2_t zn, svuint8_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_u8_x2,,,)(zn, zm);
@ -77,25 +81,27 @@ svuint8x2_t test_svadd_vector_single2_u8(svuint8x2_t zn, svuint8_t zm) __arm_str
// CHECK-LABEL: @test_svadd_vector_single2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x2.nxv8i16(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single2_s1611svint16x2_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x2.nxv8i16(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svint16x2_t test_svadd_vector_single2_s16(svint16x2_t zn, svint16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_s16_x2,,,)(zn, zm);
@ -103,25 +109,27 @@ svint16x2_t test_svadd_vector_single2_s16(svint16x2_t zn, svint16_t zm) __arm_st
// CHECK-LABEL: @test_svadd_vector_single2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x2.nxv8i16(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single2_u1612svuint16x2_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x2.nxv8i16(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svuint16x2_t test_svadd_vector_single2_u16(svuint16x2_t zn, svuint16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_u16_x2,,,)(zn, zm);
@ -129,25 +137,27 @@ svuint16x2_t test_svadd_vector_single2_u16(svuint16x2_t zn, svuint16_t zm) __arm
// CHECK-LABEL: @test_svadd_vector_single2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single2_s3211svint32x2_tu11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svint32x2_t test_svadd_vector_single2_s32(svint32x2_t zn, svint32_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_s32_x2,,,)(zn, zm);
@ -155,25 +165,27 @@ svint32x2_t test_svadd_vector_single2_s32(svint32x2_t zn, svint32_t zm) __arm_st
// CHECK-LABEL: @test_svadd_vector_single2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single2_u3212svuint32x2_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svuint32x2_t test_svadd_vector_single2_u32(svuint32x2_t zn, svuint32_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_u32_x2,,,)(zn, zm);
@ -181,25 +193,27 @@ svuint32x2_t test_svadd_vector_single2_u32(svuint32x2_t zn, svuint32_t zm) __arm
// CHECK-LABEL: @test_svadd_vector_single2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x2.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single2_s6411svint64x2_tu11__SVInt64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x2.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
svint64x2_t test_svadd_vector_single2_s64(svint64x2_t zn, svint64_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_s64_x2,,,)(zn, zm);
@ -207,25 +221,27 @@ svint64x2_t test_svadd_vector_single2_s64(svint64x2_t zn, svint64_t zm) __arm_st
// CHECK-LABEL: @test_svadd_vector_single2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP6]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x2.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single2_u6412svuint64x2_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP6]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x2.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
svuint64x2_t test_svadd_vector_single2_u64(svuint64x2_t zn, svuint64_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_u64_x2,,,)(zn, zm);
@ -236,37 +252,35 @@ svuint64x2_t test_svadd_vector_single2_u64(svuint64x2_t zn, svuint64_t zm) __arm
// CHECK-LABEL: @test_svadd_vector_single4_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]], i64 32)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x4.nxv16i8(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z28test_svadd_vector_single4_s810svint8x4_tu10__SVInt8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]], i64 32)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x4.nxv16i8(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CPP-CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
svint8x4_t test_svadd_vector_single4_s8(svint8x4_t zn, svint8_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_s8_x4,,,)(zn, zm);
@ -274,37 +288,35 @@ svint8x4_t test_svadd_vector_single4_s8(svint8x4_t zn, svint8_t zm) __arm_stream
// CHECK-LABEL: @test_svadd_vector_single4_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]], i64 32)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x4.nxv16i8(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z28test_svadd_vector_single4_u811svuint8x4_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN]], i64 48)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]], i64 32)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.add.single.x4.nxv16i8(<vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZN_COERCE2:%.*]], <vscale x 16 x i8> [[ZN_COERCE3:%.*]], <vscale x 16 x i8> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CPP-CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
svuint8x4_t test_svadd_vector_single4_u8(svuint8x4_t zn, svuint8_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_u8_x4,,,)(zn, zm);
@ -312,37 +324,35 @@ svuint8x4_t test_svadd_vector_single4_u8(svuint8x4_t zn, svuint8_t zm) __arm_str
// CHECK-LABEL: @test_svadd_vector_single4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 16)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x4.nxv8i16(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single4_s1611svint16x4_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 16)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x4.nxv8i16(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svint16x4_t test_svadd_vector_single4_s16(svint16x4_t zn, svint16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_s16_x4,,,)(zn, zm);
@ -350,37 +360,35 @@ svint16x4_t test_svadd_vector_single4_s16(svint16x4_t zn, svint16_t zm) __arm_st
// CHECK-LABEL: @test_svadd_vector_single4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 16)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x4.nxv8i16(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single4_u1612svuint16x4_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 16)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.add.single.x4.nxv8i16(<vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZN_COERCE2:%.*]], <vscale x 8 x i16> [[ZN_COERCE3:%.*]], <vscale x 8 x i16> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svuint16x4_t test_svadd_vector_single4_u16(svuint16x4_t zn, svuint16_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_u16_x4,,,)(zn, zm);
@ -388,37 +396,35 @@ svuint16x4_t test_svadd_vector_single4_u16(svuint16x4_t zn, svuint16_t zm) __arm
// CHECK-LABEL: @test_svadd_vector_single4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single4_s3211svint32x4_tu11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svint32x4_t test_svadd_vector_single4_s32(svint32x4_t zn, svint32_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_s32_x4,,,)(zn, zm);
@ -426,37 +432,35 @@ svint32x4_t test_svadd_vector_single4_s32(svint32x4_t zn, svint32_t zm) __arm_st
// CHECK-LABEL: @test_svadd_vector_single4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 8)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single4_u3212svuint32x4_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 8)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.add.single.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], <vscale x 4 x i32> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svuint32x4_t test_svadd_vector_single4_u32(svuint32x4_t zn, svuint32_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_u32_x4,,,)(zn, zm);
@ -464,37 +468,35 @@ svuint32x4_t test_svadd_vector_single4_u32(svuint32x4_t zn, svuint32_t zm) __arm
// CHECK-LABEL: @test_svadd_vector_single4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 4)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single4_s6411svint64x4_tu11__SVInt64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 4)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
svint64x4_t test_svadd_vector_single4_s64(svint64x4_t zn, svint64_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_s64_x4,,,)(zn, zm);
@ -502,37 +504,35 @@ svint64x4_t test_svadd_vector_single4_s64(svint64x4_t zn, svint64_t zm) __arm_st
// CHECK-LABEL: @test_svadd_vector_single4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 2
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 4)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 3
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP12]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z29test_svadd_vector_single4_u6412svuint64x4_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 2
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 4)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 3
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP12]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.add.single.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], <vscale x 2 x i64> [[ZM:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
svuint64x4_t test_svadd_vector_single4_u64(svuint64x4_t zn, svuint64_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svadd,_single_u64_x4,,,)(zn, zm);

228
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_vector_qrshr.c
View File

@ -20,17 +20,13 @@
// CHECK-LABEL: @test_svsqrshr_u16_u32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshr.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshr.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svsqrshr_u16_u32_x412svuint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshr.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshr.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_svsqrshr_u16_u32_x4(svuint32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshr,_n,_u16,_u32_x2,)(zn, 16);
@ -38,17 +34,13 @@ svuint16_t test_svsqrshr_u16_u32_x4(svuint32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshr_s16_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshr.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshr.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svsqrshr_s16_s32_x411svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshr.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshr.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svint16_t test_svsqrshr_s16_s32_x4(svint32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshr,_n,_s16,_s32_x2,)(zn, 16);
@ -56,21 +48,13 @@ svint16_t test_svsqrshr_s16_s32_x4(svint32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshr_u8_u32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqrshr.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 8)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqrshr.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 8)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z23test_svsqrshr_u8_u32_x412svuint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqrshr.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqrshr.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svuint8_t test_svsqrshr_u8_u32_x4(svuint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshr,_n,_u8,_u32_x4,)(zn, 8);
@ -78,21 +62,13 @@ svuint8_t test_svsqrshr_u8_u32_x4(svuint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshr_s8_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshr.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 8)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshr.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 8)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z23test_svsqrshr_s8_s32_x411svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshr.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshr.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svint8_t test_svsqrshr_s8_s32_x4(svint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshr,_n,_s8,_s32_x4,)(zn, 8);
@ -100,21 +76,13 @@ svint8_t test_svsqrshr_s8_s32_x4(svint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshr_u16_u64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshr.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshr.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svsqrshr_u16_u64_x412svuint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshr.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshr.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_svsqrshr_u16_u64_x4(svuint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshr,_n,_u16,_u64_x4,)(zn, 16);
@ -122,21 +90,13 @@ svuint16_t test_svsqrshr_u16_u64_x4(svuint64x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshr_s16_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshr.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshr.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svsqrshr_s16_s64_x411svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshr.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshr.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svint16_t test_svsqrshr_s16_s64_x4(svint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshr,_n,_s16,_s64_x4,)(zn, 16);
@ -146,21 +106,13 @@ svint16_t test_svsqrshr_s16_s64_x4(svint64x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshrn_u8_u32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqrshrn.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 8)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqrshrn.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 8)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svsqrshrn_u8_u32_x412svuint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqrshrn.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.uqrshrn.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svuint8_t test_svsqrshrn_u8_u32_x4(svuint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshrn,_n,_u8,_u32_x4,)(zn, 8);
@ -168,21 +120,13 @@ svuint8_t test_svsqrshrn_u8_u32_x4(svuint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshrn_s8_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshrn.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 8)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshrn.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 8)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svsqrshrn_s8_s32_x411svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshrn.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshrn.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svint8_t test_svsqrshrn_s8_s32_x4(svint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshrn,_n,_s8,_s32_x4,)(zn, 8);
@ -190,21 +134,13 @@ svint8_t test_svsqrshrn_s8_s32_x4(svint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshrn_u16_u64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshrn.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshrn.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z25test_svsqrshrn_u16_u64_x412svuint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshrn.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshrn.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_svsqrshrn_u16_u64_x4(svuint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshrn,_n,_u16,_u64_x4,)(zn, 16);
@ -212,21 +148,13 @@ svuint16_t test_svsqrshrn_u16_u64_x4(svuint64x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshrn_s16_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrn.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrn.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z25test_svsqrshrn_s16_s64_x411svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrn.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrn.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svint16_t test_svsqrshrn_s16_s64_x4(svint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshrn,_n,_s16,_s64_x4,)(zn, 16);
@ -236,17 +164,13 @@ svint16_t test_svsqrshrn_s16_s64_x4(svint64x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsvqrshru_u16_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshru.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshru.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z26test_svsvqrshru_u16_s32_x211svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshru.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshru.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_svsvqrshru_u16_s32_x2(svint32x2_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshru,_n,_u16,_s32_x2,)(zn, 16);
@ -254,21 +178,13 @@ svuint16_t test_svsvqrshru_u16_s32_x2(svint32x2_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshru_u8_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshru.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 8)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshru.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 8)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svsqrshru_u8_s32_x411svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshru.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshru.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svuint8_t test_svsqrshru_u8_s32_x4(svint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshru,_n,_u8,_s32_x4,)(zn, 8);
@ -276,21 +192,13 @@ svuint8_t test_svsqrshru_u8_s32_x4(svint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshru_u16_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshru.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshru.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z25test_svsqrshru_u16_s64_x411svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshru.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshru.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_svsqrshru_u16_s64_x4(svint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshru,_n,_u16,_s64_x4,)(zn, 16);
@ -300,21 +208,13 @@ svuint16_t test_svsqrshru_u16_s64_x4(svint64x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshrun_u8_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshrun.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 32)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshrun.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 32)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z25test_svsqrshrun_u8_s32_x411svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshrun.x4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i32 32)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.sqrshrun.x4.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZN_COERCE2:%.*]], <vscale x 4 x i32> [[ZN_COERCE3:%.*]], i32 32)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svuint8_t test_svsqrshrun_u8_s32_x4(svint32x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshrun,_n,_u8,_s32_x4,)(zn, 32);
@ -322,21 +222,13 @@ svuint8_t test_svsqrshrun_u8_s32_x4(svint32x4_t zn) __arm_streaming {
// CHECK-LABEL: @test_svsqrshrun_u16_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrun.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 64)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrun.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 64)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z26test_svsqrshrun_u16_s64_x411svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrun.x4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i32 64)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrun.x4.nxv2i64(<vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZN_COERCE2:%.*]], <vscale x 2 x i64> [[ZN_COERCE3:%.*]], i32 64)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_svsqrshrun_u16_s64_x4(svint64x4_t zn) __arm_streaming {
return SVE_ACLE_FUNC(svqrshrun,_n,_u16,_s64_x4,)(zn, 64);

1504
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_vector_rshl.c
View File

File diff suppressed because it is too large Load Diff

456
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_vector_selx2.c
View File

@ -19,29 +19,27 @@
// CHECK-LABEL: @test_svsel_s8_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x2.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x2.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM_COERCE0:%.*]], <vscale x 16 x i8> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z16test_svsel_s8_x2u11__SVCount_t10svint8x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x2.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x2.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM_COERCE0:%.*]], <vscale x 16 x i8> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
svint8x2_t test_svsel_s8_x2(svcount_t pn, svint8x2_t zn, svint8x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_s8_x2)(pn, zn, zm);
@ -49,29 +47,27 @@ svint8x2_t test_svsel_s8_x2(svcount_t pn, svint8x2_t zn, svint8x2_t zm) __arm_st
// CHECK-LABEL: @test_svsel_u8_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM]], i64 16)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x2.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x2.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM_COERCE0:%.*]], <vscale x 16 x i8> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z16test_svsel_u8_x2u11__SVCount_t11svuint8x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZN]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[ZM]], i64 16)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x2.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x2.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[ZN_COERCE0:%.*]], <vscale x 16 x i8> [[ZN_COERCE1:%.*]], <vscale x 16 x i8> [[ZM_COERCE0:%.*]], <vscale x 16 x i8> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: store <vscale x 32 x i8> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP5]]
//
svuint8x2_t test_svsel_u8_x2(svcount_t pn, svuint8x2_t zn, svuint8x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_u8_x2)(pn, zn, zm);
@ -81,29 +77,27 @@ svuint8x2_t test_svsel_u8_x2(svcount_t pn, svuint8x2_t zn, svuint8x2_t zm) __arm
// CHECK-LABEL: @test_svsel_s16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x2.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x2.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_s16_x2u11__SVCount_t11svint16x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x2.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x2.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svint16x2_t test_svsel_s16_x2(svcount_t pn, svint16x2_t zn, svint16x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_s16_x2)(pn, zn, zm);
@ -111,29 +105,27 @@ svint16x2_t test_svsel_s16_x2(svcount_t pn, svint16x2_t zn, svint16x2_t zm) __ar
// CHECK-LABEL: @test_svsel_u16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x2.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x2.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_u16_x2u11__SVCount_t12svuint16x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x2.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x2.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[ZN_COERCE0:%.*]], <vscale x 8 x i16> [[ZN_COERCE1:%.*]], <vscale x 8 x i16> [[ZM_COERCE0:%.*]], <vscale x 8 x i16> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x i16> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP5]]
//
svuint16x2_t test_svsel_u16_x2(svcount_t pn, svuint16x2_t zn, svuint16x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_u16_x2)(pn, zn, zm);
@ -141,29 +133,27 @@ svuint16x2_t test_svsel_u16_x2(svcount_t pn, svuint16x2_t zn, svuint16x2_t zm) _
// CHECK-LABEL: @test_svsel_f16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.sel.x2.nxv8f16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.sel.x2.nxv8f16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_f16_x2u11__SVCount_t13svfloat16x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.sel.x2.nxv8f16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.sel.x2.nxv8f16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x half> [[ZN_COERCE0:%.*]], <vscale x 8 x half> [[ZN_COERCE1:%.*]], <vscale x 8 x half> [[ZM_COERCE0:%.*]], <vscale x 8 x half> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x half> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP5]]
//
svfloat16x2_t test_svsel_f16_x2(svcount_t pn, svfloat16x2_t zn, svfloat16x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_f16_x2)(pn, zn, zm);
@ -171,29 +161,27 @@ svfloat16x2_t test_svsel_f16_x2(svcount_t pn, svfloat16x2_t zn, svfloat16x2_t zm
// CHECK-LABEL: @test_svsel_bf16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.sel.x2.nxv8bf16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.sel.x2.nxv8bf16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z18test_svsel_bf16_x2u11__SVCount_t14svbfloat16x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZN]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[ZM]], i64 8)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.sel.x2.nxv8bf16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.sel.x2.nxv8bf16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: store <vscale x 16 x bfloat> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP5]]
//
svbfloat16x2_t test_svsel_bf16_x2(svcount_t pn, svbfloat16x2_t zn, svbfloat16x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_bf16_x2)(pn, zn, zm);
@ -203,29 +191,27 @@ svbfloat16x2_t test_svsel_bf16_x2(svcount_t pn, svbfloat16x2_t zn, svbfloat16x2_
// CHECK-LABEL: @test_svsel_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x2.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x2.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_s32_x2u11__SVCount_t11svint32x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x2.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x2.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svint32x2_t test_svsel_s32_x2(svcount_t pn, svint32x2_t zn, svint32x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_s32_x2)(pn, zn, zm);
@ -233,29 +219,27 @@ svint32x2_t test_svsel_s32_x2(svcount_t pn, svint32x2_t zn, svint32x2_t zm) __ar
// CHECK-LABEL: @test_svsel_u32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x2.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x2.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_u32_x2u11__SVCount_t12svuint32x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x2.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x2.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], <vscale x 4 x i32> [[ZM_COERCE0:%.*]], <vscale x 4 x i32> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x i32> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP5]]
//
svuint32x2_t test_svsel_u32_x2(svcount_t pn, svuint32x2_t zn, svuint32x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_u32_x2)(pn, zn, zm);
@ -263,29 +247,27 @@ svuint32x2_t test_svsel_u32_x2(svcount_t pn, svuint32x2_t zn, svuint32x2_t zm) _
// CHECK-LABEL: @test_svsel_f32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.sel.x2.nxv4f32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.sel.x2.nxv4f32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_f32_x2u11__SVCount_t13svfloat32x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[ZM]], i64 4)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.sel.x2.nxv4f32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.sel.x2.nxv4f32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x float> [[ZN_COERCE0:%.*]], <vscale x 4 x float> [[ZN_COERCE1:%.*]], <vscale x 4 x float> [[ZM_COERCE0:%.*]], <vscale x 4 x float> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: store <vscale x 8 x float> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP5]]
//
svfloat32x2_t test_svsel_f32_x2(svcount_t pn, svfloat32x2_t zn, svfloat32x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_f32_x2)(pn, zn, zm);
@ -295,29 +277,27 @@ svfloat32x2_t test_svsel_f32_x2(svcount_t pn, svfloat32x2_t zn, svfloat32x2_t zm
// CHECK-LABEL: @test_svsel_s64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x2.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x2.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_s64_x2u11__SVCount_t11svint64x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x2.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x2.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
svint64x2_t test_svsel_s64_x2(svcount_t pn, svint64x2_t zn, svint64x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_s64_x2)(pn, zn, zm);
@ -325,29 +305,27 @@ svint64x2_t test_svsel_s64_x2(svcount_t pn, svint64x2_t zn, svint64x2_t zm) __ar
// CHECK-LABEL: @test_svsel_u64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x2.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x2.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_u64_x2u11__SVCount_t12svuint64x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x2.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x2.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[ZN_COERCE0:%.*]], <vscale x 2 x i64> [[ZN_COERCE1:%.*]], <vscale x 2 x i64> [[ZM_COERCE0:%.*]], <vscale x 2 x i64> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x i64> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP5]]
//
svuint64x2_t test_svsel_u64_x2(svcount_t pn, svuint64x2_t zn, svuint64x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_u64_x2)(pn, zn, zm);
@ -355,29 +333,27 @@ svuint64x2_t test_svsel_u64_x2(svcount_t pn, svuint64x2_t zn, svuint64x2_t zm) _
// CHECK-LABEL: @test_svsel_f64_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.sel.x2.nxv2f64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x double> [[TMP8]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.sel.x2.nxv2f64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_f64_x2u11__SVCount_t13svfloat64x2_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZN]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[ZM]], i64 2)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.sel.x2.nxv2f64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 0
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP5]], i64 0)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]], 1
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x double> [[TMP8]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.sel.x2.nxv2f64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x double> [[ZN_COERCE0:%.*]], <vscale x 2 x double> [[ZN_COERCE1:%.*]], <vscale x 2 x double> [[ZM_COERCE0:%.*]], <vscale x 2 x double> [[ZM_COERCE1:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: store <vscale x 4 x double> [[TMP4]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP5]]
//
svfloat64x2_t test_svsel_f64_x2(svcount_t pn, svfloat64x2_t zn, svfloat64x2_t zm) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_f64_x2)(pn, zn, zm);

744
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_vector_selx4.c
View File

@ -19,45 +19,35 @@
// CHECK-LABEL: @test_svsel_s8_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 48)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 16)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 48)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x4.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 16)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP13]], i64 32)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x4.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[ZN1_COERCE0:%.*]], <vscale x 16 x i8> [[ZN1_COERCE1:%.*]], <vscale x 16 x i8> [[ZN1_COERCE2:%.*]], <vscale x 16 x i8> [[ZN1_COERCE3:%.*]], <vscale x 16 x i8> [[ZN2_COERCE0:%.*]], <vscale x 16 x i8> [[ZN2_COERCE1:%.*]], <vscale x 16 x i8> [[ZN2_COERCE2:%.*]], <vscale x 16 x i8> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z16test_svsel_s8_x4u11__SVCount_t10svint8x4_tS0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 48)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 16)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 48)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x4.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 16)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP13]], i64 32)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x4.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[ZN1_COERCE0:%.*]], <vscale x 16 x i8> [[ZN1_COERCE1:%.*]], <vscale x 16 x i8> [[ZN1_COERCE2:%.*]], <vscale x 16 x i8> [[ZN1_COERCE3:%.*]], <vscale x 16 x i8> [[ZN2_COERCE0:%.*]], <vscale x 16 x i8> [[ZN2_COERCE1:%.*]], <vscale x 16 x i8> [[ZN2_COERCE2:%.*]], <vscale x 16 x i8> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CPP-CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
svint8x4_t test_svsel_s8_x4(svcount_t pn, svint8x4_t zn1, svint8x4_t zn2) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_s8_x4)(pn, zn1, zn2);
@ -65,45 +55,35 @@ svint8x4_t test_svsel_s8_x4(svcount_t pn, svint8x4_t zn1, svint8x4_t zn2) __arm_
// CHECK-LABEL: @test_svsel_u8_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 48)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 16)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 48)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x4.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 16)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP13]], i64 32)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x4.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[ZN1_COERCE0:%.*]], <vscale x 16 x i8> [[ZN1_COERCE1:%.*]], <vscale x 16 x i8> [[ZN1_COERCE2:%.*]], <vscale x 16 x i8> [[ZN1_COERCE3:%.*]], <vscale x 16 x i8> [[ZN2_COERCE0:%.*]], <vscale x 16 x i8> [[ZN2_COERCE1:%.*]], <vscale x 16 x i8> [[ZN2_COERCE2:%.*]], <vscale x 16 x i8> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z16test_svsel_u8_x4u11__SVCount_t11svuint8x4_tS0_S0_S0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN1]], i64 48)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 16)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[ZN2]], i64 48)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x4.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], i64 16)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP13]], i64 32)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.sel.x4.nxv16i8(target("aarch64.svcount") [[PN:%.*]], <vscale x 16 x i8> [[ZN1_COERCE0:%.*]], <vscale x 16 x i8> [[ZN1_COERCE1:%.*]], <vscale x 16 x i8> [[ZN1_COERCE2:%.*]], <vscale x 16 x i8> [[ZN1_COERCE3:%.*]], <vscale x 16 x i8> [[ZN2_COERCE0:%.*]], <vscale x 16 x i8> [[ZN2_COERCE1:%.*]], <vscale x 16 x i8> [[ZN2_COERCE2:%.*]], <vscale x 16 x i8> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CPP-CHECK-NEXT: store <vscale x 64 x i8> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP9]]
//
svuint8x4_t test_svsel_u8_x4(svcount_t pn, svuint8x4_t zn1, svuint8x4_t zn2, svuint8x4_t zn3, svuint8x4_t zn4) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_u8_x4)(pn, zn1, zn2);
@ -113,45 +93,35 @@ svuint8x4_t test_svsel_u8_x4(svcount_t pn, svuint8x4_t zn1, svuint8x4_t zn2, svu
// CHECK-LABEL: @test_svsel_s16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 24)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x4.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 8)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP12]], <vscale x 8 x i16> [[TMP13]], i64 16)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP14]], <vscale x 8 x i16> [[TMP15]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x4.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[ZN1_COERCE0:%.*]], <vscale x 8 x i16> [[ZN1_COERCE1:%.*]], <vscale x 8 x i16> [[ZN1_COERCE2:%.*]], <vscale x 8 x i16> [[ZN1_COERCE3:%.*]], <vscale x 8 x i16> [[ZN2_COERCE0:%.*]], <vscale x 8 x i16> [[ZN2_COERCE1:%.*]], <vscale x 8 x i16> [[ZN2_COERCE2:%.*]], <vscale x 8 x i16> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_s16_x4u11__SVCount_t11svint16x4_tS0_S0_S0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 24)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x4.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 8)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP12]], <vscale x 8 x i16> [[TMP13]], i64 16)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP14]], <vscale x 8 x i16> [[TMP15]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x4.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[ZN1_COERCE0:%.*]], <vscale x 8 x i16> [[ZN1_COERCE1:%.*]], <vscale x 8 x i16> [[ZN1_COERCE2:%.*]], <vscale x 8 x i16> [[ZN1_COERCE3:%.*]], <vscale x 8 x i16> [[ZN2_COERCE0:%.*]], <vscale x 8 x i16> [[ZN2_COERCE1:%.*]], <vscale x 8 x i16> [[ZN2_COERCE2:%.*]], <vscale x 8 x i16> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svint16x4_t test_svsel_s16_x4(svcount_t pn, svint16x4_t zn1, svint16x4_t zn2, svint16x4_t zn3, svint16x4_t zn4) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_s16_x4)(pn, zn1, zn2);
@ -159,45 +129,35 @@ svint16x4_t test_svsel_s16_x4(svcount_t pn, svint16x4_t zn1, svint16x4_t zn2, sv
// CHECK-LABEL: @test_svsel_u16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 24)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x4.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 8)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP12]], <vscale x 8 x i16> [[TMP13]], i64 16)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP14]], <vscale x 8 x i16> [[TMP15]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x4.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[ZN1_COERCE0:%.*]], <vscale x 8 x i16> [[ZN1_COERCE1:%.*]], <vscale x 8 x i16> [[ZN1_COERCE2:%.*]], <vscale x 8 x i16> [[ZN1_COERCE3:%.*]], <vscale x 8 x i16> [[ZN2_COERCE0:%.*]], <vscale x 8 x i16> [[ZN2_COERCE1:%.*]], <vscale x 8 x i16> [[ZN2_COERCE2:%.*]], <vscale x 8 x i16> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_u16_x4u11__SVCount_t12svuint16x4_tS0_S0_S0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN1]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[ZN2]], i64 24)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x4.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP10]], <vscale x 8 x i16> [[TMP11]], i64 8)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP12]], <vscale x 8 x i16> [[TMP13]], i64 16)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP14]], <vscale x 8 x i16> [[TMP15]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.sel.x4.nxv8i16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x i16> [[ZN1_COERCE0:%.*]], <vscale x 8 x i16> [[ZN1_COERCE1:%.*]], <vscale x 8 x i16> [[ZN1_COERCE2:%.*]], <vscale x 8 x i16> [[ZN1_COERCE3:%.*]], <vscale x 8 x i16> [[ZN2_COERCE0:%.*]], <vscale x 8 x i16> [[ZN2_COERCE1:%.*]], <vscale x 8 x i16> [[ZN2_COERCE2:%.*]], <vscale x 8 x i16> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x i16> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP9]]
//
svuint16x4_t test_svsel_u16_x4(svcount_t pn, svuint16x4_t zn1, svuint16x4_t zn2, svuint16x4_t zn3, svuint16x4_t zn4) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_u16_x4)(pn, zn1, zn2);
@ -205,45 +165,35 @@ svuint16x4_t test_svsel_u16_x4(svcount_t pn, svuint16x4_t zn1, svuint16x4_t zn2,
// CHECK-LABEL: @test_svsel_f16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN1]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN1]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN1]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN2]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN2]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN2]], i64 24)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.sel.x4.nxv8f16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 8)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP12]], <vscale x 8 x half> [[TMP13]], i64 16)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP14]], <vscale x 8 x half> [[TMP15]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.sel.x4.nxv8f16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x half> [[ZN1_COERCE0:%.*]], <vscale x 8 x half> [[ZN1_COERCE1:%.*]], <vscale x 8 x half> [[ZN1_COERCE2:%.*]], <vscale x 8 x half> [[ZN1_COERCE3:%.*]], <vscale x 8 x half> [[ZN2_COERCE0:%.*]], <vscale x 8 x half> [[ZN2_COERCE1:%.*]], <vscale x 8 x half> [[ZN2_COERCE2:%.*]], <vscale x 8 x half> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_f16_x4u11__SVCount_t13svfloat16x4_tS0_S0_S0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN1]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN1]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN1]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN2]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN2]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN2]], i64 24)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.sel.x4.nxv8f16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP10]], <vscale x 8 x half> [[TMP11]], i64 8)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP12]], <vscale x 8 x half> [[TMP13]], i64 16)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP14]], <vscale x 8 x half> [[TMP15]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.sel.x4.nxv8f16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x half> [[ZN1_COERCE0:%.*]], <vscale x 8 x half> [[ZN1_COERCE1:%.*]], <vscale x 8 x half> [[ZN1_COERCE2:%.*]], <vscale x 8 x half> [[ZN1_COERCE3:%.*]], <vscale x 8 x half> [[ZN2_COERCE0:%.*]], <vscale x 8 x half> [[ZN2_COERCE1:%.*]], <vscale x 8 x half> [[ZN2_COERCE2:%.*]], <vscale x 8 x half> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x half> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP9]]
//
svfloat16x4_t test_svsel_f16_x4(svcount_t pn, svfloat16x4_t zn1, svfloat16x4_t zn2, svfloat16x4_t zn3, svfloat16x4_t zn4) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_f16_x4)(pn, zn1, zn2);
@ -251,45 +201,35 @@ svfloat16x4_t test_svsel_f16_x4(svcount_t pn, svfloat16x4_t zn1, svfloat16x4_t z
// CHECK-LABEL: @test_svsel_bf16_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN1]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN1]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN1]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN2]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN2]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN2]], i64 24)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.sel.x4.nxv8bf16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP10]], <vscale x 8 x bfloat> [[TMP11]], i64 8)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP12]], <vscale x 8 x bfloat> [[TMP13]], i64 16)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP14]], <vscale x 8 x bfloat> [[TMP15]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.sel.x4.nxv8bf16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x bfloat> [[ZN1_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN1_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN1_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN1_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZN2_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN2_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN2_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z18test_svsel_bf16_x4u11__SVCount_t14svbfloat16x4_tS0_S0_S0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN1]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN1]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN1]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN2]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN2]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[ZN2]], i64 24)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.sel.x4.nxv8bf16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP10]], <vscale x 8 x bfloat> [[TMP11]], i64 8)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP12]], <vscale x 8 x bfloat> [[TMP13]], i64 16)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP14]], <vscale x 8 x bfloat> [[TMP15]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.sel.x4.nxv8bf16(target("aarch64.svcount") [[PN:%.*]], <vscale x 8 x bfloat> [[ZN1_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN1_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN1_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN1_COERCE3:%.*]], <vscale x 8 x bfloat> [[ZN2_COERCE0:%.*]], <vscale x 8 x bfloat> [[ZN2_COERCE1:%.*]], <vscale x 8 x bfloat> [[ZN2_COERCE2:%.*]], <vscale x 8 x bfloat> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], i64 8)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 24)
// CPP-CHECK-NEXT: store <vscale x 32 x bfloat> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP9]]
//
svbfloat16x4_t test_svsel_bf16_x4(svcount_t pn, svbfloat16x4_t zn1, svbfloat16x4_t zn2, svbfloat16x4_t zn3, svbfloat16x4_t zn4) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_bf16_x4)(pn, zn1, zn2);
@ -299,45 +239,35 @@ svbfloat16x4_t test_svsel_bf16_x4(svcount_t pn, svbfloat16x4_t zn1, svbfloat16x4
// CHECK-LABEL: @test_svsel_s32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 12)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x4.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 4)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP12]], <vscale x 4 x i32> [[TMP13]], i64 8)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP14]], <vscale x 4 x i32> [[TMP15]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x4.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[ZN1_COERCE0:%.*]], <vscale x 4 x i32> [[ZN1_COERCE1:%.*]], <vscale x 4 x i32> [[ZN1_COERCE2:%.*]], <vscale x 4 x i32> [[ZN1_COERCE3:%.*]], <vscale x 4 x i32> [[ZN2_COERCE0:%.*]], <vscale x 4 x i32> [[ZN2_COERCE1:%.*]], <vscale x 4 x i32> [[ZN2_COERCE2:%.*]], <vscale x 4 x i32> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_s32_x4u11__SVCount_t11svint32x4_tS0_S0_S0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 12)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x4.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 4)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP12]], <vscale x 4 x i32> [[TMP13]], i64 8)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP14]], <vscale x 4 x i32> [[TMP15]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x4.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[ZN1_COERCE0:%.*]], <vscale x 4 x i32> [[ZN1_COERCE1:%.*]], <vscale x 4 x i32> [[ZN1_COERCE2:%.*]], <vscale x 4 x i32> [[ZN1_COERCE3:%.*]], <vscale x 4 x i32> [[ZN2_COERCE0:%.*]], <vscale x 4 x i32> [[ZN2_COERCE1:%.*]], <vscale x 4 x i32> [[ZN2_COERCE2:%.*]], <vscale x 4 x i32> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svint32x4_t test_svsel_s32_x4(svcount_t pn, svint32x4_t zn1, svint32x4_t zn2, svint32x4_t zn3, svint32x4_t zn4) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_s32_x4)(pn, zn1, zn2);
@ -345,45 +275,35 @@ svint32x4_t test_svsel_s32_x4(svcount_t pn, svint32x4_t zn1, svint32x4_t zn2, sv
// CHECK-LABEL: @test_svsel_u32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 12)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x4.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 4)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP12]], <vscale x 4 x i32> [[TMP13]], i64 8)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP14]], <vscale x 4 x i32> [[TMP15]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x4.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[ZN1_COERCE0:%.*]], <vscale x 4 x i32> [[ZN1_COERCE1:%.*]], <vscale x 4 x i32> [[ZN1_COERCE2:%.*]], <vscale x 4 x i32> [[ZN1_COERCE3:%.*]], <vscale x 4 x i32> [[ZN2_COERCE0:%.*]], <vscale x 4 x i32> [[ZN2_COERCE1:%.*]], <vscale x 4 x i32> [[ZN2_COERCE2:%.*]], <vscale x 4 x i32> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_u32_x4u11__SVCount_t12svuint32x4_tS0_S0_S0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN1]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[ZN2]], i64 12)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x4.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP10]], <vscale x 4 x i32> [[TMP11]], i64 4)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP12]], <vscale x 4 x i32> [[TMP13]], i64 8)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP14]], <vscale x 4 x i32> [[TMP15]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.sel.x4.nxv4i32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x i32> [[ZN1_COERCE0:%.*]], <vscale x 4 x i32> [[ZN1_COERCE1:%.*]], <vscale x 4 x i32> [[ZN1_COERCE2:%.*]], <vscale x 4 x i32> [[ZN1_COERCE3:%.*]], <vscale x 4 x i32> [[ZN2_COERCE0:%.*]], <vscale x 4 x i32> [[ZN2_COERCE1:%.*]], <vscale x 4 x i32> [[ZN2_COERCE2:%.*]], <vscale x 4 x i32> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x i32> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP9]]
//
svuint32x4_t test_svsel_u32_x4(svcount_t pn, svuint32x4_t zn1, svuint32x4_t zn2, svuint32x4_t zn3, svuint32x4_t zn4) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_u32_x4)(pn, zn1, zn2);
@ -391,45 +311,35 @@ svuint32x4_t test_svsel_u32_x4(svcount_t pn, svuint32x4_t zn1, svuint32x4_t zn2,
// CHECK-LABEL: @test_svsel_f32_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN1]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN1]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN1]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN2]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN2]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN2]], i64 12)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.sel.x4.nxv4f32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 4)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP12]], <vscale x 4 x float> [[TMP13]], i64 8)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP14]], <vscale x 4 x float> [[TMP15]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.sel.x4.nxv4f32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x float> [[ZN1_COERCE0:%.*]], <vscale x 4 x float> [[ZN1_COERCE1:%.*]], <vscale x 4 x float> [[ZN1_COERCE2:%.*]], <vscale x 4 x float> [[ZN1_COERCE3:%.*]], <vscale x 4 x float> [[ZN2_COERCE0:%.*]], <vscale x 4 x float> [[ZN2_COERCE1:%.*]], <vscale x 4 x float> [[ZN2_COERCE2:%.*]], <vscale x 4 x float> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_f32_x4u11__SVCount_t13svfloat32x4_tS0_S0_S0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN1]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN1]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN1]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN2]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN2]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[ZN2]], i64 12)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.sel.x4.nxv4f32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP10]], <vscale x 4 x float> [[TMP11]], i64 4)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP12]], <vscale x 4 x float> [[TMP13]], i64 8)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP14]], <vscale x 4 x float> [[TMP15]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.sel.x4.nxv4f32(target("aarch64.svcount") [[PN:%.*]], <vscale x 4 x float> [[ZN1_COERCE0:%.*]], <vscale x 4 x float> [[ZN1_COERCE1:%.*]], <vscale x 4 x float> [[ZN1_COERCE2:%.*]], <vscale x 4 x float> [[ZN1_COERCE3:%.*]], <vscale x 4 x float> [[ZN2_COERCE0:%.*]], <vscale x 4 x float> [[ZN2_COERCE1:%.*]], <vscale x 4 x float> [[ZN2_COERCE2:%.*]], <vscale x 4 x float> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], i64 4)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 12)
// CPP-CHECK-NEXT: store <vscale x 16 x float> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP9]]
//
svfloat32x4_t test_svsel_f32_x4(svcount_t pn, svfloat32x4_t zn1, svfloat32x4_t zn2, svfloat32x4_t zn3, svfloat32x4_t zn4) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_f32_x4)(pn, zn1, zn2);
@ -439,45 +349,35 @@ svfloat32x4_t test_svsel_f32_x4(svcount_t pn, svfloat32x4_t zn1, svfloat32x4_t z
// CHECK-LABEL: @test_svsel_s64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 6)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x4.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 2)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP12]], <vscale x 2 x i64> [[TMP13]], i64 4)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP14]], <vscale x 2 x i64> [[TMP15]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x4.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[ZN1_COERCE0:%.*]], <vscale x 2 x i64> [[ZN1_COERCE1:%.*]], <vscale x 2 x i64> [[ZN1_COERCE2:%.*]], <vscale x 2 x i64> [[ZN1_COERCE3:%.*]], <vscale x 2 x i64> [[ZN2_COERCE0:%.*]], <vscale x 2 x i64> [[ZN2_COERCE1:%.*]], <vscale x 2 x i64> [[ZN2_COERCE2:%.*]], <vscale x 2 x i64> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_s64_x4u11__SVCount_t11svint64x4_tS0_S0_S0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 6)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x4.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 2)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP12]], <vscale x 2 x i64> [[TMP13]], i64 4)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP14]], <vscale x 2 x i64> [[TMP15]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x4.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[ZN1_COERCE0:%.*]], <vscale x 2 x i64> [[ZN1_COERCE1:%.*]], <vscale x 2 x i64> [[ZN1_COERCE2:%.*]], <vscale x 2 x i64> [[ZN1_COERCE3:%.*]], <vscale x 2 x i64> [[ZN2_COERCE0:%.*]], <vscale x 2 x i64> [[ZN2_COERCE1:%.*]], <vscale x 2 x i64> [[ZN2_COERCE2:%.*]], <vscale x 2 x i64> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
svint64x4_t test_svsel_s64_x4(svcount_t pn, svint64x4_t zn1, svint64x4_t zn2, svint64x4_t zn3, svint64x4_t zn4) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_s64_x4)(pn, zn1, zn2);
@ -485,45 +385,35 @@ svint64x4_t test_svsel_s64_x4(svcount_t pn, svint64x4_t zn1, svint64x4_t zn2, sv
// CHECK-LABEL: @test_svsel_u64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 6)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x4.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 2)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP12]], <vscale x 2 x i64> [[TMP13]], i64 4)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP14]], <vscale x 2 x i64> [[TMP15]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x4.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[ZN1_COERCE0:%.*]], <vscale x 2 x i64> [[ZN1_COERCE1:%.*]], <vscale x 2 x i64> [[ZN1_COERCE2:%.*]], <vscale x 2 x i64> [[ZN1_COERCE3:%.*]], <vscale x 2 x i64> [[ZN2_COERCE0:%.*]], <vscale x 2 x i64> [[ZN2_COERCE1:%.*]], <vscale x 2 x i64> [[ZN2_COERCE2:%.*]], <vscale x 2 x i64> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_u64_x4u11__SVCount_t12svuint64x4_tS0_S0_S0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN1]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[ZN2]], i64 6)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x4.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP10]], <vscale x 2 x i64> [[TMP11]], i64 2)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP12]], <vscale x 2 x i64> [[TMP13]], i64 4)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP14]], <vscale x 2 x i64> [[TMP15]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.sel.x4.nxv2i64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x i64> [[ZN1_COERCE0:%.*]], <vscale x 2 x i64> [[ZN1_COERCE1:%.*]], <vscale x 2 x i64> [[ZN1_COERCE2:%.*]], <vscale x 2 x i64> [[ZN1_COERCE3:%.*]], <vscale x 2 x i64> [[ZN2_COERCE0:%.*]], <vscale x 2 x i64> [[ZN2_COERCE1:%.*]], <vscale x 2 x i64> [[ZN2_COERCE2:%.*]], <vscale x 2 x i64> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x i64> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP9]]
//
svuint64x4_t test_svsel_u64_x4(svcount_t pn, svuint64x4_t zn1, svuint64x4_t zn2, svuint64x4_t zn3, svuint64x4_t zn4) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_u64_x4)(pn, zn1, zn2);
@ -531,45 +421,35 @@ svuint64x4_t test_svsel_u64_x4(svcount_t pn, svuint64x4_t zn1, svuint64x4_t zn2,
// CHECK-LABEL: @test_svsel_f64_x4(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN1:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN1]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN1]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN1]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN2:%.*]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN2]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN2]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN2]], i64 6)
// CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.sel.x4.nxv2f64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 0
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP9]], i64 0)
// CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 1
// CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 2)
// CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 2
// CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP12]], <vscale x 2 x double> [[TMP13]], i64 4)
// CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 3
// CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP14]], <vscale x 2 x double> [[TMP15]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x double> [[TMP16]]
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.sel.x4.nxv2f64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x double> [[ZN1_COERCE0:%.*]], <vscale x 2 x double> [[ZN1_COERCE1:%.*]], <vscale x 2 x double> [[ZN1_COERCE2:%.*]], <vscale x 2 x double> [[ZN1_COERCE3:%.*]], <vscale x 2 x double> [[ZN2_COERCE0:%.*]], <vscale x 2 x double> [[ZN2_COERCE1:%.*]], <vscale x 2 x double> [[ZN2_COERCE2:%.*]], <vscale x 2 x double> [[ZN2_COERCE3:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
// CPP-CHECK-LABEL: @_Z17test_svsel_f64_x4u11__SVCount_t13svfloat64x4_tS0_S0_S0_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN1:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN1]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN1]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN1]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN2:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN2]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN2]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[ZN2]], i64 6)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.sel.x4.nxv2f64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 0
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP9]], i64 0)
// CPP-CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 1
// CPP-CHECK-NEXT: [[TMP12:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP10]], <vscale x 2 x double> [[TMP11]], i64 2)
// CPP-CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 2
// CPP-CHECK-NEXT: [[TMP14:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP12]], <vscale x 2 x double> [[TMP13]], i64 4)
// CPP-CHECK-NEXT: [[TMP15:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP8]], 3
// CPP-CHECK-NEXT: [[TMP16:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP14]], <vscale x 2 x double> [[TMP15]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x double> [[TMP16]]
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, align 16
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.sel.x4.nxv2f64(target("aarch64.svcount") [[PN:%.*]], <vscale x 2 x double> [[ZN1_COERCE0:%.*]], <vscale x 2 x double> [[ZN1_COERCE1:%.*]], <vscale x 2 x double> [[ZN1_COERCE2:%.*]], <vscale x 2 x double> [[ZN1_COERCE3:%.*]], <vscale x 2 x double> [[ZN2_COERCE0:%.*]], <vscale x 2 x double> [[ZN2_COERCE1:%.*]], <vscale x 2 x double> [[ZN2_COERCE2:%.*]], <vscale x 2 x double> [[ZN2_COERCE3:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], i64 2)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 6)
// CPP-CHECK-NEXT: store <vscale x 8 x double> [[TMP8]], ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: [[TMP9:%.*]] = load { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }, ptr [[RETVAL]], align 16
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP9]]
//
svfloat64x4_t test_svsel_f64_x4(svcount_t pn, svfloat64x4_t zn1, svfloat64x4_t zn2, svfloat64x4_t zn3, svfloat64x4_t zn4) __arm_streaming {
return SVE_ACLE_FUNC(svsel,_f64_x4)(pn, zn1, zn2);

816
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_vector_uzpx2.c
View File

File diff suppressed because it is too large Load Diff

1296
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_vector_uzpx4.c
View File

File diff suppressed because it is too large Load Diff

816
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_vector_zipx2.c
View File

File diff suppressed because it is too large Load Diff

1296
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_vector_zipx4.c
View File

File diff suppressed because it is too large Load Diff

720
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_write.c
View File

File diff suppressed because it is too large Load Diff

1016
clang/test/CodeGen/aarch64-sme2p1-intrinsics/acle_sme2p1_movaz.c
View File

File diff suppressed because it is too large Load Diff

12
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create2-bfloat.c
View File

@ -24,15 +24,15 @@
// CHECK-LABEL: @test_svcreate2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z19test_svcreate2_bf16u14__SVBfloat16_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
//
svbfloat16x2_t test_svcreate2_bf16(svbfloat16_t x0, svbfloat16_t x1) ATTR
{

132
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create2.c
View File

@ -24,15 +24,15 @@
// CHECK-LABEL: @test_svcreate2_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z17test_svcreate2_s8u10__SVInt8_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
svint8x2_t test_svcreate2_s8(svint8_t x0, svint8_t x1) ATTR
{
@ -41,15 +41,15 @@ svint8x2_t test_svcreate2_s8(svint8_t x0, svint8_t x1) ATTR
// CHECK-LABEL: @test_svcreate2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate2_s16u11__SVInt16_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
svint16x2_t test_svcreate2_s16(svint16_t x0, svint16_t x1) ATTR
{
@ -58,15 +58,15 @@ svint16x2_t test_svcreate2_s16(svint16_t x0, svint16_t x1) ATTR
// CHECK-LABEL: @test_svcreate2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate2_s32u11__SVInt32_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
svint32x2_t test_svcreate2_s32(svint32_t x0, svint32_t x1) ATTR
{
@ -75,15 +75,15 @@ svint32x2_t test_svcreate2_s32(svint32_t x0, svint32_t x1) ATTR
// CHECK-LABEL: @test_svcreate2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate2_s64u11__SVInt64_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
svint64x2_t test_svcreate2_s64(svint64_t x0, svint64_t x1) ATTR
{
@ -92,15 +92,15 @@ svint64x2_t test_svcreate2_s64(svint64_t x0, svint64_t x1) ATTR
// CHECK-LABEL: @test_svcreate2_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z17test_svcreate2_u8u11__SVUint8_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
svuint8x2_t test_svcreate2_u8(svuint8_t x0, svuint8_t x1) ATTR
{
@ -109,15 +109,15 @@ svuint8x2_t test_svcreate2_u8(svuint8_t x0, svuint8_t x1) ATTR
// CHECK-LABEL: @test_svcreate2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate2_u16u12__SVUint16_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
svuint16x2_t test_svcreate2_u16(svuint16_t x0, svuint16_t x1) ATTR
{
@ -126,15 +126,15 @@ svuint16x2_t test_svcreate2_u16(svuint16_t x0, svuint16_t x1) ATTR
// CHECK-LABEL: @test_svcreate2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate2_u32u12__SVUint32_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
svuint32x2_t test_svcreate2_u32(svuint32_t x0, svuint32_t x1) ATTR
{
@ -143,15 +143,15 @@ svuint32x2_t test_svcreate2_u32(svuint32_t x0, svuint32_t x1) ATTR
// CHECK-LABEL: @test_svcreate2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate2_u64u12__SVUint64_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
svuint64x2_t test_svcreate2_u64(svuint64_t x0, svuint64_t x1) ATTR
{
@ -160,15 +160,15 @@ svuint64x2_t test_svcreate2_u64(svuint64_t x0, svuint64_t x1) ATTR
// CHECK-LABEL: @test_svcreate2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate2_f16u13__SVFloat16_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
//
svfloat16x2_t test_svcreate2_f16(svfloat16_t x0, svfloat16_t x1) ATTR
{
@ -177,15 +177,15 @@ svfloat16x2_t test_svcreate2_f16(svfloat16_t x0, svfloat16_t x1) ATTR
// CHECK-LABEL: @test_svcreate2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate2_f32u13__SVFloat32_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
//
svfloat32x2_t test_svcreate2_f32(svfloat32_t x0, svfloat32_t x1) ATTR
{
@ -194,15 +194,15 @@ svfloat32x2_t test_svcreate2_f32(svfloat32_t x0, svfloat32_t x1) ATTR
// CHECK-LABEL: @test_svcreate2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x double> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate2_f64u13__SVFloat64_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x double> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
//
svfloat64x2_t test_svcreate2_f64(svfloat64_t x0, svfloat64_t x1) ATTR
{

16
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create3-bfloat.c
View File

@ -24,17 +24,17 @@
// CHECK-LABEL: @test_svcreate3_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svcreate3_bf16u14__SVBfloat16_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
svbfloat16x3_t test_svcreate3_bf16(svbfloat16_t x0, svbfloat16_t x1, svbfloat16_t x2) ATTR
{

176
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create3.c
View File

@ -24,17 +24,17 @@
// CHECK-LABEL: @test_svcreate3_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
// CHECK-NEXT: ret <vscale x 48 x i8> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z17test_svcreate3_s8u10__SVInt8_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
// CPP-CHECK-NEXT: ret <vscale x 48 x i8> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]]
//
svint8x3_t test_svcreate3_s8(svint8_t x0, svint8_t x1, svint8_t x2) ATTR
{
@ -43,17 +43,17 @@ svint8x3_t test_svcreate3_s8(svint8_t x0, svint8_t x1, svint8_t x2) ATTR
// CHECK-LABEL: @test_svcreate3_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate3_s16u11__SVInt16_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
svint16x3_t test_svcreate3_s16(svint16_t x0, svint16_t x1, svint16_t x2) ATTR
{
@ -62,17 +62,17 @@ svint16x3_t test_svcreate3_s16(svint16_t x0, svint16_t x1, svint16_t x2) ATTR
// CHECK-LABEL: @test_svcreate3_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 12 x i32> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate3_s32u11__SVInt32_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 12 x i32> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
svint32x3_t test_svcreate3_s32(svint32_t x0, svint32_t x1, svint32_t x2) ATTR
{
@ -81,17 +81,17 @@ svint32x3_t test_svcreate3_s32(svint32_t x0, svint32_t x1, svint32_t x2) ATTR
// CHECK-LABEL: @test_svcreate3_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 6 x i64> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate3_s64u11__SVInt64_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 6 x i64> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
svint64x3_t test_svcreate3_s64(svint64_t x0, svint64_t x1, svint64_t x2) ATTR
{
@ -100,17 +100,17 @@ svint64x3_t test_svcreate3_s64(svint64_t x0, svint64_t x1, svint64_t x2) ATTR
// CHECK-LABEL: @test_svcreate3_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
// CHECK-NEXT: ret <vscale x 48 x i8> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z17test_svcreate3_u8u11__SVUint8_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
// CPP-CHECK-NEXT: ret <vscale x 48 x i8> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]]
//
svuint8x3_t test_svcreate3_u8(svuint8_t x0, svuint8_t x1, svuint8_t x2) ATTR
{
@ -119,17 +119,17 @@ svuint8x3_t test_svcreate3_u8(svuint8_t x0, svuint8_t x1, svuint8_t x2) ATTR
// CHECK-LABEL: @test_svcreate3_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate3_u16u12__SVUint16_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
svuint16x3_t test_svcreate3_u16(svuint16_t x0, svuint16_t x1, svuint16_t x2) ATTR
{
@ -138,17 +138,17 @@ svuint16x3_t test_svcreate3_u16(svuint16_t x0, svuint16_t x1, svuint16_t x2) ATT
// CHECK-LABEL: @test_svcreate3_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 12 x i32> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate3_u32u12__SVUint32_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 12 x i32> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
svuint32x3_t test_svcreate3_u32(svuint32_t x0, svuint32_t x1, svuint32_t x2) ATTR
{
@ -157,17 +157,17 @@ svuint32x3_t test_svcreate3_u32(svuint32_t x0, svuint32_t x1, svuint32_t x2) ATT
// CHECK-LABEL: @test_svcreate3_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 6 x i64> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate3_u64u12__SVUint64_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 6 x i64> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
svuint64x3_t test_svcreate3_u64(svuint64_t x0, svuint64_t x1, svuint64_t x2) ATTR
{
@ -176,17 +176,17 @@ svuint64x3_t test_svcreate3_u64(svuint64_t x0, svuint64_t x1, svuint64_t x2) ATT
// CHECK-LABEL: @test_svcreate3_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP1]], <vscale x 8 x half> [[X2:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x half> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate3_f16u13__SVFloat16_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP1]], <vscale x 8 x half> [[X2:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x half> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]]
//
svfloat16x3_t test_svcreate3_f16(svfloat16_t x0, svfloat16_t x1, svfloat16_t x2) ATTR
{
@ -195,17 +195,17 @@ svfloat16x3_t test_svcreate3_f16(svfloat16_t x0, svfloat16_t x1, svfloat16_t x2)
// CHECK-LABEL: @test_svcreate3_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP1]], <vscale x 4 x float> [[X2:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 12 x float> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate3_f32u13__SVFloat32_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP1]], <vscale x 4 x float> [[X2:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 12 x float> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]]
//
svfloat32x3_t test_svcreate3_f32(svfloat32_t x0, svfloat32_t x1, svfloat32_t x2) ATTR
{
@ -214,17 +214,17 @@ svfloat32x3_t test_svcreate3_f32(svfloat32_t x0, svfloat32_t x1, svfloat32_t x2)
// CHECK-LABEL: @test_svcreate3_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP1]], <vscale x 2 x double> [[X2:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 6 x double> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[X2:%.*]], 2
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate3_f64u13__SVFloat64_tS_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP1]], <vscale x 2 x double> [[X2:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 6 x double> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[X2:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]]
//
svfloat64x3_t test_svcreate3_f64(svfloat64_t x0, svfloat64_t x1, svfloat64_t x2) ATTR
{

20
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create4-bfloat.c
View File

@ -24,19 +24,19 @@
// CHECK-LABEL: @test_svcreate4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[X4:%.*]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z19test_svcreate4_bf16u14__SVBfloat16_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[X4:%.*]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]]
//
svbfloat16x4_t test_svcreate4_bf16(svbfloat16_t x0, svbfloat16_t x1, svbfloat16_t x2, svbfloat16_t x4) ATTR
{

220
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create4.c
View File

@ -24,19 +24,19 @@
// CHECK-LABEL: @test_svcreate4_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z17test_svcreate4_s8u10__SVInt8_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
//
svint8x4_t test_svcreate4_s8(svint8_t x0, svint8_t x1, svint8_t x2, svint8_t x4) ATTR
{
@ -45,19 +45,19 @@ svint8x4_t test_svcreate4_s8(svint8_t x0, svint8_t x1, svint8_t x2, svint8_t x4)
// CHECK-LABEL: @test_svcreate4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate4_s16u11__SVInt16_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]]
//
svint16x4_t test_svcreate4_s16(svint16_t x0, svint16_t x1, svint16_t x2, svint16_t x4) ATTR
{
@ -66,19 +66,19 @@ svint16x4_t test_svcreate4_s16(svint16_t x0, svint16_t x1, svint16_t x2, svint16
// CHECK-LABEL: @test_svcreate4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate4_s32u11__SVInt32_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]]
//
svint32x4_t test_svcreate4_s32(svint32_t x0, svint32_t x1, svint32_t x2, svint32_t x4) ATTR
{
@ -87,19 +87,19 @@ svint32x4_t test_svcreate4_s32(svint32_t x0, svint32_t x1, svint32_t x2, svint32
// CHECK-LABEL: @test_svcreate4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate4_s64u11__SVInt64_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]]
//
svint64x4_t test_svcreate4_s64(svint64_t x0, svint64_t x1, svint64_t x2, svint64_t x4) ATTR
{
@ -108,19 +108,19 @@ svint64x4_t test_svcreate4_s64(svint64_t x0, svint64_t x1, svint64_t x2, svint64
// CHECK-LABEL: @test_svcreate4_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z17test_svcreate4_u8u11__SVUint8_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
//
svuint8x4_t test_svcreate4_u8(svuint8_t x0, svuint8_t x1, svuint8_t x2, svuint8_t x4) ATTR
{
@ -129,19 +129,19 @@ svuint8x4_t test_svcreate4_u8(svuint8_t x0, svuint8_t x1, svuint8_t x2, svuint8_
// CHECK-LABEL: @test_svcreate4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate4_u16u12__SVUint16_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]]
//
svuint16x4_t test_svcreate4_u16(svuint16_t x0, svuint16_t x1, svuint16_t x2, svuint16_t x4) ATTR
{
@ -150,19 +150,19 @@ svuint16x4_t test_svcreate4_u16(svuint16_t x0, svuint16_t x1, svuint16_t x2, svu
// CHECK-LABEL: @test_svcreate4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate4_u32u12__SVUint32_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]]
//
svuint32x4_t test_svcreate4_u32(svuint32_t x0, svuint32_t x1, svuint32_t x2, svuint32_t x4) ATTR
{
@ -171,19 +171,19 @@ svuint32x4_t test_svcreate4_u32(svuint32_t x0, svuint32_t x1, svuint32_t x2, svu
// CHECK-LABEL: @test_svcreate4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate4_u64u12__SVUint64_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]]
//
svuint64x4_t test_svcreate4_u64(svuint64_t x0, svuint64_t x1, svuint64_t x2, svuint64_t x4) ATTR
{
@ -192,19 +192,19 @@ svuint64x4_t test_svcreate4_u64(svuint64_t x0, svuint64_t x1, svuint64_t x2, svu
// CHECK-LABEL: @test_svcreate4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP1]], <vscale x 8 x half> [[X2:%.*]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[X4:%.*]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate4_f16u13__SVFloat16_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP1]], <vscale x 8 x half> [[X2:%.*]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[X4:%.*]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]]
//
svfloat16x4_t test_svcreate4_f16(svfloat16_t x0, svfloat16_t x1, svfloat16_t x2, svfloat16_t x4) ATTR
{
@ -213,19 +213,19 @@ svfloat16x4_t test_svcreate4_f16(svfloat16_t x0, svfloat16_t x1, svfloat16_t x2,
// CHECK-LABEL: @test_svcreate4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP1]], <vscale x 4 x float> [[X2:%.*]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[X4:%.*]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate4_f32u13__SVFloat32_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP1]], <vscale x 4 x float> [[X2:%.*]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[X4:%.*]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]]
//
svfloat32x4_t test_svcreate4_f32(svfloat32_t x0, svfloat32_t x1, svfloat32_t x2, svfloat32_t x4) ATTR
{
@ -234,19 +234,19 @@ svfloat32x4_t test_svcreate4_f32(svfloat32_t x0, svfloat32_t x1, svfloat32_t x2,
// CHECK-LABEL: @test_svcreate4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP1]], <vscale x 2 x double> [[X2:%.*]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[X4:%.*]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x double> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svcreate4_f64u13__SVFloat64_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP1]], <vscale x 2 x double> [[X2:%.*]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[X4:%.*]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x double> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]]
//
svfloat64x4_t test_svcreate4_f64(svfloat64_t x0, svfloat64_t x1, svfloat64_t x2, svfloat64_t x4) ATTR
{

24
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get2-bfloat.c
View File

@ -24,13 +24,17 @@
// CHECK-LABEL: @test_svget2_bf16_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svget2_bf16_014svbfloat16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
//
svbfloat16_t test_svget2_bf16_0(svbfloat16x2_t tuple) ATTR
{
@ -39,13 +43,17 @@ svbfloat16_t test_svget2_bf16_0(svbfloat16x2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_bf16_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svget2_bf16_114svbfloat16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
//
svbfloat16_t test_svget2_bf16_1(svbfloat16x2_t tuple) ATTR
{

132
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get2.c
View File

@ -25,13 +25,17 @@
// CHECK-LABEL: @test_svget2_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z14test_svget2_s810svint8x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP2]]
//
svint8_t test_svget2_s8(svint8x2_t tuple) ATTR
{
@ -40,13 +44,17 @@ svint8_t test_svget2_s8(svint8x2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svget2_s1611svint16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
//
svint16_t test_svget2_s16(svint16x2_t tuple) ATTR
{
@ -55,13 +63,17 @@ svint16_t test_svget2_s16(svint16x2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svget2_s3211svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
//
svint32_t test_svget2_s32(svint32x2_t tuple) ATTR
{
@ -70,13 +82,17 @@ svint32_t test_svget2_s32(svint32x2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i64 2)
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svget2_s6411svint64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
//
svint64_t test_svget2_s64(svint64x2_t tuple) ATTR
{
@ -85,13 +101,17 @@ svint64_t test_svget2_s64(svint64x2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z14test_svget2_u811svuint8x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP2]]
//
svuint8_t test_svget2_u8(svuint8x2_t tuple) ATTR
{
@ -100,13 +120,17 @@ svuint8_t test_svget2_u8(svuint8x2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svget2_u1612svuint16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
//
svuint16_t test_svget2_u16(svuint16x2_t tuple) ATTR
{
@ -115,13 +139,17 @@ svuint16_t test_svget2_u16(svuint16x2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svget2_u3212svuint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
//
svuint32_t test_svget2_u32(svuint32x2_t tuple) ATTR
{
@ -130,13 +158,17 @@ svuint32_t test_svget2_u32(svuint32x2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i64 2)
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svget2_u6412svuint64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
//
svuint64_t test_svget2_u64(svuint64x2_t tuple) ATTR
{
@ -145,13 +177,17 @@ svuint64_t test_svget2_u64(svuint64x2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svget2_f1613svfloat16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP2]]
//
svfloat16_t test_svget2_f16(svfloat16x2_t tuple) ATTR
{
@ -160,13 +196,17 @@ svfloat16_t test_svget2_f16(svfloat16x2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[TUPLE:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CHECK-NEXT: ret <vscale x 4 x float> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svget2_f3213svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[TUPLE:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP2]]
//
svfloat32_t test_svget2_f32(svfloat32x2_t tuple) ATTR
{
@ -175,13 +215,17 @@ svfloat32_t test_svget2_f32(svfloat32x2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 2 x double> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CHECK-NEXT: ret <vscale x 2 x double> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svget2_f6413svfloat64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 2 x double> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CPP-CHECK-NEXT: ret <vscale x 2 x double> [[TMP2]]
//
svfloat64_t test_svget2_f64(svfloat64x2_t tuple) ATTR
{

42
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get3-bfloat.c
View File

@ -24,13 +24,19 @@
// CHECK-LABEL: @test_svget3_bf16_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svget3_bf16_014svbfloat16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP3]]
//
svbfloat16_t test_svget3_bf16_0(svbfloat16x3_t tuple) ATTR
{
@ -39,13 +45,19 @@ svbfloat16_t test_svget3_bf16_0(svbfloat16x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_bf16_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svget3_bf16_114svbfloat16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP3]]
//
svbfloat16_t test_svget3_bf16_1(svbfloat16x3_t tuple) ATTR
{
@ -54,13 +66,19 @@ svbfloat16_t test_svget3_bf16_1(svbfloat16x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_bf16_2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 2
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svget3_bf16_214svbfloat16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 2
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP3]]
//
svbfloat16_t test_svget3_bf16_2(svbfloat16x3_t tuple) ATTR
{

154
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get3.c
View File

@ -26,13 +26,19 @@
// CHECK-LABEL: @test_svget3_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z14test_svget3_s810svint8x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP3]]
//
svint8_t test_svget3_s8(svint8x3_t tuple) ATTR
{
@ -41,13 +47,19 @@ svint8_t test_svget3_s8(svint8x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svget3_s1611svint16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP3]]
//
svint16_t test_svget3_s16(svint16x3_t tuple) ATTR
{
@ -56,13 +68,19 @@ svint16_t test_svget3_s16(svint16x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svget3_s3211svint32x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP3]]
//
svint32_t test_svget3_s32(svint32x3_t tuple) ATTR
{
@ -71,13 +89,19 @@ svint32_t test_svget3_s32(svint32x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svget3_s6411svint64x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP3]]
//
svint64_t test_svget3_s64(svint64x3_t tuple) ATTR
{
@ -86,13 +110,19 @@ svint64_t test_svget3_s64(svint64x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i64 32)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z14test_svget3_u811svuint8x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i64 32)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP3]]
//
svuint8_t test_svget3_u8(svuint8x3_t tuple) ATTR
{
@ -101,13 +131,19 @@ svuint8_t test_svget3_u8(svuint8x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svget3_u1612svuint16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP3]]
//
svuint16_t test_svget3_u16(svuint16x3_t tuple) ATTR
{
@ -116,13 +152,19 @@ svuint16_t test_svget3_u16(svuint16x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svget3_u3212svuint32x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP3]]
//
svuint32_t test_svget3_u32(svuint32x3_t tuple) ATTR
{
@ -131,13 +173,19 @@ svuint32_t test_svget3_u32(svuint32x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svget3_u6412svuint64x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP3]]
//
svuint64_t test_svget3_u64(svuint64x3_t tuple) ATTR
{
@ -146,13 +194,19 @@ svuint64_t test_svget3_u64(svuint64x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[TUPLE:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svget3_f1613svfloat16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[TUPLE:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP3]]
//
svfloat16_t test_svget3_f16(svfloat16x3_t tuple) ATTR
{
@ -161,13 +215,19 @@ svfloat16_t test_svget3_f16(svfloat16x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: ret <vscale x 4 x float> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svget3_f3213svfloat32x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP3]]
//
svfloat32_t test_svget3_f32(svfloat32x3_t tuple) ATTR
{
@ -176,13 +236,19 @@ svfloat32_t test_svget3_f32(svfloat32x3_t tuple) ATTR
// CHECK-LABEL: @test_svget3_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[TUPLE:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 2 x double> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 2
// CHECK-NEXT: ret <vscale x 2 x double> [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svget3_f6413svfloat64x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[TUPLE:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 2 x double> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 2
// CPP-CHECK-NEXT: ret <vscale x 2 x double> [[TMP3]]
//
svfloat64_t test_svget3_f64(svfloat64x3_t tuple) ATTR
{

64
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get4-bfloat.c
View File

@ -24,13 +24,21 @@
// CHECK-LABEL: @test_svget4_bf16_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z18test_svget4_bf16_014svbfloat16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP4]]
//
svbfloat16_t test_svget4_bf16_0(svbfloat16x4_t tuple) ATTR
{
@ -39,13 +47,21 @@ svbfloat16_t test_svget4_bf16_0(svbfloat16x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_bf16_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z18test_svget4_bf16_114svbfloat16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP4]]
//
svbfloat16_t test_svget4_bf16_1(svbfloat16x4_t tuple) ATTR
{
@ -54,13 +70,21 @@ svbfloat16_t test_svget4_bf16_1(svbfloat16x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_bf16_2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 2
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z18test_svget4_bf16_214svbfloat16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 2
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP4]]
//
svbfloat16_t test_svget4_bf16_2(svbfloat16x4_t tuple) ATTR
{
@ -69,13 +93,21 @@ svbfloat16_t test_svget4_bf16_2(svbfloat16x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_bf16_3(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 24)
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 3
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z18test_svget4_bf16_314svbfloat16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 3
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP4]]
//
svbfloat16_t test_svget4_bf16_3(svbfloat16x4_t tuple) ATTR
{

176
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get4.c
View File

@ -27,13 +27,21 @@
// several parameters, one for each member of the original struct.
// CHECK-LABEL: @test_svget4_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 0
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z14test_svget4_s810svint8x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 0
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
//
svint8_t test_svget4_s8(svint8x4_t tuple) ATTR
{
@ -42,13 +50,21 @@ svint8_t test_svget4_s8(svint8x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_s1611svint16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
//
svint16_t test_svget4_s16(svint16x4_t tuple) ATTR
{
@ -57,13 +73,21 @@ svint16_t test_svget4_s16(svint16x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_s3211svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP4]]
//
svint32_t test_svget4_s32(svint32x4_t tuple) ATTR
{
@ -72,13 +96,21 @@ svint32_t test_svget4_s32(svint32x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i64 6)
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_s6411svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP4]]
//
svint64_t test_svget4_s64(svint64x4_t tuple) ATTR
{
@ -87,13 +119,21 @@ svint64_t test_svget4_s64(svint64x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i64 32)
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 2
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z14test_svget4_u811svuint8x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i64 32)
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 2
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP4]]
//
svuint8_t test_svget4_u8(svuint8x4_t tuple) ATTR
{
@ -102,13 +142,21 @@ svuint8_t test_svget4_u8(svuint8x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i64 24)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_u1612svuint16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP4]]
//
svuint16_t test_svget4_u16(svuint16x4_t tuple) ATTR
{
@ -117,13 +165,21 @@ svuint16_t test_svget4_u16(svuint16x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_u3212svuint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP4]]
//
svuint32_t test_svget4_u32(svuint32x4_t tuple) ATTR
{
@ -132,13 +188,21 @@ svuint32_t test_svget4_u32(svuint32x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i64 6)
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_u6412svuint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP4]]
//
svuint64_t test_svget4_u64(svuint64x4_t tuple) ATTR
{
@ -147,13 +211,21 @@ svuint64_t test_svget4_u64(svuint64x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[TUPLE:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 2
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_f1613svfloat16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[TUPLE:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 2
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP4]]
//
svfloat16_t test_svget4_f16(svfloat16x4_t tuple) ATTR
{
@ -162,13 +234,21 @@ svfloat16_t test_svget4_f16(svfloat16x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
// CHECK-NEXT: ret <vscale x 4 x float> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_f3213svfloat32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP4]]
//
svfloat32_t test_svget4_f32(svfloat32x4_t tuple) ATTR
{
@ -177,13 +257,21 @@ svfloat32_t test_svget4_f32(svfloat32x4_t tuple) ATTR
// CHECK-LABEL: @test_svget4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[TUPLE:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 2 x double> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 2
// CHECK-NEXT: ret <vscale x 2 x double> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_f6413svfloat64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[TUPLE:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 2 x double> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 2
// CPP-CHECK-NEXT: ret <vscale x 2 x double> [[TMP4]]
//
svfloat64_t test_svget4_f64(svfloat64x4_t tuple) ATTR
{

24
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld2-bfloat.c
View File

@ -26,21 +26,13 @@
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z15test_svld2_bf16u10__SVBool_tPKu6__bf16(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
//
svbfloat16x2_t test_svld2_bf16(svbool_t pg, const bfloat16_t *base) MODE_ATTR
{
@ -53,22 +45,14 @@ svbfloat16x2_t test_svld2_bf16(svbool_t pg, const bfloat16_t *base) MODE_ATTR
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z20test_svld2_vnum_bf16u10__SVBool_tPKu6__bf16l(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
svbfloat16x2_t test_svld2_vnum_bf16(svbool_t pg, const bfloat16_t *base, int64_t vnum) MODE_ATTR
{

264
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld2.c
View File

@ -26,20 +26,12 @@
// CHECK-LABEL: @test_svld2_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z13test_svld2_s8u10__SVBool_tPKa(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
svint8x2_t test_svld2_s8(svbool_t pg, const int8_t *base) MODE_ATTR
{
@ -50,21 +42,13 @@ svint8x2_t test_svld2_s8(svbool_t pg, const int8_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld2_s16u10__SVBool_tPKs(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
svint16x2_t test_svld2_s16(svbool_t pg, const int16_t *base) MODE_ATTR
{
@ -75,21 +59,13 @@ svint16x2_t test_svld2_s16(svbool_t pg, const int16_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld2_s32u10__SVBool_tPKi(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
svint32x2_t test_svld2_s32(svbool_t pg, const int32_t *base) MODE_ATTR
{
@ -100,21 +76,13 @@ svint32x2_t test_svld2_s32(svbool_t pg, const int32_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld2_s64u10__SVBool_tPKl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
svint64x2_t test_svld2_s64(svbool_t pg, const int64_t *base) MODE_ATTR
{
@ -124,20 +92,12 @@ svint64x2_t test_svld2_s64(svbool_t pg, const int64_t *base) MODE_ATTR
// CHECK-LABEL: @test_svld2_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z13test_svld2_u8u10__SVBool_tPKh(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP4]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
svuint8x2_t test_svld2_u8(svbool_t pg, const uint8_t *base) MODE_ATTR
{
@ -148,21 +108,13 @@ svuint8x2_t test_svld2_u8(svbool_t pg, const uint8_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld2_u16u10__SVBool_tPKt(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
svuint16x2_t test_svld2_u16(svbool_t pg, const uint16_t *base) MODE_ATTR
{
@ -173,21 +125,13 @@ svuint16x2_t test_svld2_u16(svbool_t pg, const uint16_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld2_u32u10__SVBool_tPKj(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
svuint32x2_t test_svld2_u32(svbool_t pg, const uint32_t *base) MODE_ATTR
{
@ -198,21 +142,13 @@ svuint32x2_t test_svld2_u32(svbool_t pg, const uint32_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld2_u64u10__SVBool_tPKm(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
svuint64x2_t test_svld2_u64(svbool_t pg, const uint64_t *base) MODE_ATTR
{
@ -223,21 +159,13 @@ svuint64x2_t test_svld2_u64(svbool_t pg, const uint64_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld2_f16u10__SVBool_tPKDh(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
//
svfloat16x2_t test_svld2_f16(svbool_t pg, const float16_t *base) MODE_ATTR
{
@ -248,21 +176,13 @@ svfloat16x2_t test_svld2_f16(svbool_t pg, const float16_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld2_f32u10__SVBool_tPKf(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
//
svfloat32x2_t test_svld2_f32(svbool_t pg, const float32_t *base) MODE_ATTR
{
@ -273,21 +193,13 @@ svfloat32x2_t test_svld2_f32(svbool_t pg, const float32_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x double> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld2_f64u10__SVBool_tPKd(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x double> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
//
svfloat64x2_t test_svld2_f64(svbool_t pg, const float64_t *base) MODE_ATTR
{
@ -298,21 +210,13 @@ svfloat64x2_t test_svld2_f64(svbool_t pg, const float64_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svld2_vnum_s8u10__SVBool_tPKal(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
svint8x2_t test_svld2_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum) MODE_ATTR
{
@ -324,22 +228,14 @@ svint8x2_t test_svld2_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum) MOD
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld2_vnum_s16u10__SVBool_tPKsl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
svint16x2_t test_svld2_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum) MODE_ATTR
{
@ -351,22 +247,14 @@ svint16x2_t test_svld2_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum)
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld2_vnum_s32u10__SVBool_tPKil(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
svint32x2_t test_svld2_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum) MODE_ATTR
{
@ -378,22 +266,14 @@ svint32x2_t test_svld2_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum)
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld2_vnum_s64u10__SVBool_tPKll(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
svint64x2_t test_svld2_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum) MODE_ATTR
{
@ -404,21 +284,13 @@ svint64x2_t test_svld2_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum)
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP5]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svld2_vnum_u8u10__SVBool_tPKhl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP5]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
svuint8x2_t test_svld2_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum) MODE_ATTR
{
@ -430,22 +302,14 @@ svuint8x2_t test_svld2_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum) M
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld2_vnum_u16u10__SVBool_tPKtl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
svuint16x2_t test_svld2_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum) MODE_ATTR
{
@ -457,22 +321,14 @@ svuint16x2_t test_svld2_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld2_vnum_u32u10__SVBool_tPKjl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
svuint32x2_t test_svld2_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum) MODE_ATTR
{
@ -484,22 +340,14 @@ svuint32x2_t test_svld2_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld2_vnum_u64u10__SVBool_tPKml(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
svuint64x2_t test_svld2_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum) MODE_ATTR
{
@ -511,22 +359,14 @@ svuint64x2_t test_svld2_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld2_vnum_f16u10__SVBool_tPKDhl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]]
//
svfloat16x2_t test_svld2_vnum_f16(svbool_t pg, const float16_t *base, int64_t vnum) MODE_ATTR
{
@ -538,22 +378,14 @@ svfloat16x2_t test_svld2_vnum_f16(svbool_t pg, const float16_t *base, int64_t vn
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld2_vnum_f32u10__SVBool_tPKfl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]]
//
svfloat32x2_t test_svld2_vnum_f32(svbool_t pg, const float32_t *base, int64_t vnum) MODE_ATTR
{
@ -565,22 +397,14 @@ svfloat32x2_t test_svld2_vnum_f32(svbool_t pg, const float32_t *base, int64_t vn
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x double> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld2_vnum_f64u10__SVBool_tPKdl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x double> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]]
//
svfloat64x2_t test_svld2_vnum_f64(svbool_t pg, const float64_t *base, int64_t vnum) MODE_ATTR
{

32
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld3-bfloat.c
View File

@ -27,25 +27,13 @@
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z15test_svld3_bf16u10__SVBool_tPKu6__bf16(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
//
svbfloat16x3_t test_svld3_bf16(svbool_t pg, const bfloat16_t *base) MODE_ATTR
{
@ -57,26 +45,14 @@ svbfloat16x3_t test_svld3_bf16(svbool_t pg, const bfloat16_t *base) MODE_ATTR
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z20test_svld3_vnum_bf16u10__SVBool_tPKu6__bf16l(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
svbfloat16x3_t test_svld3_vnum_bf16(svbool_t pg, const bfloat16_t *base, int64_t vnum) MODE_ATTR
{

352
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld3.c
View File

@ -25,24 +25,12 @@
// CHECK-LABEL: @test_svld3_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: ret <vscale x 48 x i8> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z13test_svld3_s8u10__SVBool_tPKa(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: ret <vscale x 48 x i8> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
svint8x3_t test_svld3_s8(svbool_t pg, const int8_t *base) MODE_ATTR
{
@ -53,25 +41,13 @@ svint8x3_t test_svld3_s8(svbool_t pg, const int8_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x i16> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld3_s16u10__SVBool_tPKs(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x i16> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
svint16x3_t test_svld3_s16(svbool_t pg, const int16_t *base) MODE_ATTR
{
@ -82,25 +58,13 @@ svint16x3_t test_svld3_s16(svbool_t pg, const int16_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
// CHECK-NEXT: ret <vscale x 12 x i32> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld3_s32u10__SVBool_tPKi(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 12 x i32> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
svint32x3_t test_svld3_s32(svbool_t pg, const int32_t *base) MODE_ATTR
{
@ -111,25 +75,13 @@ svint32x3_t test_svld3_s32(svbool_t pg, const int32_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
// CHECK-NEXT: ret <vscale x 6 x i64> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld3_s64u10__SVBool_tPKl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 6 x i64> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
svint64x3_t test_svld3_s64(svbool_t pg, const int64_t *base) MODE_ATTR
{
@ -139,24 +91,12 @@ svint64x3_t test_svld3_s64(svbool_t pg, const int64_t *base) MODE_ATTR
// CHECK-LABEL: @test_svld3_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: ret <vscale x 48 x i8> [[TMP6]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z13test_svld3_u8u10__SVBool_tPKh(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: ret <vscale x 48 x i8> [[TMP6]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
svuint8x3_t test_svld3_u8(svbool_t pg, const uint8_t *base) MODE_ATTR
{
@ -167,25 +107,13 @@ svuint8x3_t test_svld3_u8(svbool_t pg, const uint8_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x i16> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld3_u16u10__SVBool_tPKt(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x i16> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
svuint16x3_t test_svld3_u16(svbool_t pg, const uint16_t *base) MODE_ATTR
{
@ -196,25 +124,13 @@ svuint16x3_t test_svld3_u16(svbool_t pg, const uint16_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
// CHECK-NEXT: ret <vscale x 12 x i32> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld3_u32u10__SVBool_tPKj(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 12 x i32> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
svuint32x3_t test_svld3_u32(svbool_t pg, const uint32_t *base) MODE_ATTR
{
@ -225,25 +141,13 @@ svuint32x3_t test_svld3_u32(svbool_t pg, const uint32_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
// CHECK-NEXT: ret <vscale x 6 x i64> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld3_u64u10__SVBool_tPKm(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 6 x i64> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
svuint64x3_t test_svld3_u64(svbool_t pg, const uint64_t *base) MODE_ATTR
{
@ -254,25 +158,13 @@ svuint64x3_t test_svld3_u64(svbool_t pg, const uint64_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x half> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld3_f16u10__SVBool_tPKDh(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x half> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
//
svfloat16x3_t test_svld3_f16(svbool_t pg, const float16_t *base) MODE_ATTR
{
@ -283,25 +175,13 @@ svfloat16x3_t test_svld3_f16(svbool_t pg, const float16_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 8)
// CHECK-NEXT: ret <vscale x 12 x float> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld3_f32u10__SVBool_tPKf(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 12 x float> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
//
svfloat32x3_t test_svld3_f32(svbool_t pg, const float32_t *base) MODE_ATTR
{
@ -312,25 +192,13 @@ svfloat32x3_t test_svld3_f32(svbool_t pg, const float32_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 4)
// CHECK-NEXT: ret <vscale x 6 x double> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld3_f64u10__SVBool_tPKd(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 6 x double> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
//
svfloat64x3_t test_svld3_f64(svbool_t pg, const float64_t *base) MODE_ATTR
{
@ -341,25 +209,13 @@ svfloat64x3_t test_svld3_f64(svbool_t pg, const float64_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], i64 32)
// CHECK-NEXT: ret <vscale x 48 x i8> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svld3_vnum_s8u10__SVBool_tPKal(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], i64 32)
// CPP-CHECK-NEXT: ret <vscale x 48 x i8> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
svint8x3_t test_svld3_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum) MODE_ATTR
{
@ -371,26 +227,14 @@ svint8x3_t test_svld3_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum) MOD
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x i16> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld3_vnum_s16u10__SVBool_tPKsl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x i16> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
svint16x3_t test_svld3_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum) MODE_ATTR
{
@ -402,26 +246,14 @@ svint16x3_t test_svld3_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum)
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 12 x i32> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld3_vnum_s32u10__SVBool_tPKil(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 12 x i32> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
svint32x3_t test_svld3_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum) MODE_ATTR
{
@ -433,26 +265,14 @@ svint32x3_t test_svld3_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum)
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CHECK-NEXT: ret <vscale x 6 x i64> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld3_vnum_s64u10__SVBool_tPKll(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 6 x i64> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
svint64x3_t test_svld3_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum) MODE_ATTR
{
@ -463,25 +283,13 @@ svint64x3_t test_svld3_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum)
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], i64 32)
// CHECK-NEXT: ret <vscale x 48 x i8> [[TMP7]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svld3_vnum_u8u10__SVBool_tPKhl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], i64 32)
// CPP-CHECK-NEXT: ret <vscale x 48 x i8> [[TMP7]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
svuint8x3_t test_svld3_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum) MODE_ATTR
{
@ -493,26 +301,14 @@ svuint8x3_t test_svld3_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum) M
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x i16> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld3_vnum_u16u10__SVBool_tPKtl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x i16> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
svuint16x3_t test_svld3_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum) MODE_ATTR
{
@ -524,26 +320,14 @@ svuint16x3_t test_svld3_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 12 x i32> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld3_vnum_u32u10__SVBool_tPKjl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 12 x i32> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
svuint32x3_t test_svld3_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum) MODE_ATTR
{
@ -555,26 +339,14 @@ svuint32x3_t test_svld3_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CHECK-NEXT: ret <vscale x 6 x i64> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld3_vnum_u64u10__SVBool_tPKml(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 6 x i64> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
svuint64x3_t test_svld3_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum) MODE_ATTR
{
@ -586,26 +358,14 @@ svuint64x3_t test_svld3_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x half> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld3_vnum_f16u10__SVBool_tPKDhl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x half> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]]
//
svfloat16x3_t test_svld3_vnum_f16(svbool_t pg, const float16_t *base, int64_t vnum) MODE_ATTR
{
@ -617,26 +377,14 @@ svfloat16x3_t test_svld3_vnum_f16(svbool_t pg, const float16_t *base, int64_t vn
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 8)
// CHECK-NEXT: ret <vscale x 12 x float> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld3_vnum_f32u10__SVBool_tPKfl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 12 x float> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]]
//
svfloat32x3_t test_svld3_vnum_f32(svbool_t pg, const float32_t *base, int64_t vnum) MODE_ATTR
{
@ -648,26 +396,14 @@ svfloat32x3_t test_svld3_vnum_f32(svbool_t pg, const float32_t *base, int64_t vn
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 2)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 4)
// CHECK-NEXT: ret <vscale x 6 x double> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld3_vnum_f64u10__SVBool_tPKdl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 6 x double> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]]
//
svfloat64x3_t test_svld3_vnum_f64(svbool_t pg, const float64_t *base, int64_t vnum) MODE_ATTR
{

40
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld4-bfloat.c
View File

@ -26,29 +26,13 @@
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 16)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP7]], <vscale x 8 x bfloat> [[TMP8]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z15test_svld4_bf16u10__SVBool_tPKu6__bf16(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 16)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP7]], <vscale x 8 x bfloat> [[TMP8]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
//
svbfloat16x4_t test_svld4_bf16(svbool_t pg, const bfloat16_t *base) MODE_ATTR
{
@ -60,30 +44,14 @@ svbfloat16x4_t test_svld4_bf16(svbool_t pg, const bfloat16_t *base) MODE_ATTR
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 16)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP8]], <vscale x 8 x bfloat> [[TMP9]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP10]]
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z20test_svld4_vnum_bf16u10__SVBool_tPKu6__bf16l(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP8]], <vscale x 8 x bfloat> [[TMP9]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP10]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
svbfloat16x4_t test_svld4_vnum_bf16(svbool_t pg, const bfloat16_t *base, int64_t vnum) MODE_ATTR
{

440
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld4.c
View File

@ -25,28 +25,12 @@
// CHECK-LABEL: @test_svld4_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z13test_svld4_s8u10__SVBool_tPKa(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
svint8x4_t test_svld4_s8(svbool_t pg, const int8_t *base) MODE_ATTR
{
@ -57,29 +41,13 @@ svint8x4_t test_svld4_s8(svbool_t pg, const int8_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld4_s16u10__SVBool_tPKs(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
svint16x4_t test_svld4_s16(svbool_t pg, const int16_t *base) MODE_ATTR
{
@ -90,29 +58,13 @@ svint16x4_t test_svld4_s16(svbool_t pg, const int16_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld4_s32u10__SVBool_tPKi(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
svint32x4_t test_svld4_s32(svbool_t pg, const int32_t *base) MODE_ATTR
{
@ -123,29 +75,13 @@ svint32x4_t test_svld4_s32(svbool_t pg, const int32_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld4_s64u10__SVBool_tPKl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
svint64x4_t test_svld4_s64(svbool_t pg, const int64_t *base) MODE_ATTR
{
@ -155,28 +91,12 @@ svint64x4_t test_svld4_s64(svbool_t pg, const int64_t *base) MODE_ATTR
// CHECK-LABEL: @test_svld4_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP8]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z13test_svld4_u8u10__SVBool_tPKh(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP8]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
//
svuint8x4_t test_svld4_u8(svbool_t pg, const uint8_t *base) MODE_ATTR
{
@ -187,29 +107,13 @@ svuint8x4_t test_svld4_u8(svbool_t pg, const uint8_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld4_u16u10__SVBool_tPKt(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
//
svuint16x4_t test_svld4_u16(svbool_t pg, const uint16_t *base) MODE_ATTR
{
@ -220,29 +124,13 @@ svuint16x4_t test_svld4_u16(svbool_t pg, const uint16_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld4_u32u10__SVBool_tPKj(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
//
svuint32x4_t test_svld4_u32(svbool_t pg, const uint32_t *base) MODE_ATTR
{
@ -253,29 +141,13 @@ svuint32x4_t test_svld4_u32(svbool_t pg, const uint32_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld4_u64u10__SVBool_tPKm(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
//
svuint64x4_t test_svld4_u64(svbool_t pg, const uint64_t *base) MODE_ATTR
{
@ -286,29 +158,13 @@ svuint64x4_t test_svld4_u64(svbool_t pg, const uint64_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 16)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP7]], <vscale x 8 x half> [[TMP8]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld4_f16u10__SVBool_tPKDh(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 16)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP7]], <vscale x 8 x half> [[TMP8]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
//
svfloat16x4_t test_svld4_f16(svbool_t pg, const float16_t *base) MODE_ATTR
{
@ -319,29 +175,13 @@ svfloat16x4_t test_svld4_f16(svbool_t pg, const float16_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 8)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP7]], <vscale x 4 x float> [[TMP8]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld4_f32u10__SVBool_tPKf(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 8)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP7]], <vscale x 4 x float> [[TMP8]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
//
svfloat32x4_t test_svld4_f32(svbool_t pg, const float32_t *base) MODE_ATTR
{
@ -352,29 +192,13 @@ svfloat32x4_t test_svld4_f32(svbool_t pg, const float32_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 4)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP7]], <vscale x 2 x double> [[TMP8]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x double> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z14test_svld4_f64u10__SVBool_tPKd(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 4)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP7]], <vscale x 2 x double> [[TMP8]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x double> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
//
svfloat64x4_t test_svld4_f64(svbool_t pg, const float64_t *base) MODE_ATTR
{
@ -385,29 +209,13 @@ svfloat64x4_t test_svld4_f64(svbool_t pg, const float64_t *base) MODE_ATTR
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], i64 32)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP7]], <vscale x 16 x i8> [[TMP8]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svld4_vnum_s8u10__SVBool_tPKal(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], i64 32)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP7]], <vscale x 16 x i8> [[TMP8]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
svint8x4_t test_svld4_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum) MODE_ATTR
{
@ -419,30 +227,14 @@ svint8x4_t test_svld4_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum) MOD
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP10]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld4_vnum_s16u10__SVBool_tPKsl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP10]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
svint16x4_t test_svld4_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum) MODE_ATTR
{
@ -454,30 +246,14 @@ svint16x4_t test_svld4_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum)
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP10]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld4_vnum_s32u10__SVBool_tPKil(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP10]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
svint32x4_t test_svld4_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum) MODE_ATTR
{
@ -489,30 +265,14 @@ svint32x4_t test_svld4_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum)
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP10]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld4_vnum_s64u10__SVBool_tPKll(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP10]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
svint64x4_t test_svld4_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum) MODE_ATTR
{
@ -523,29 +283,13 @@ svint64x4_t test_svld4_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum)
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 2
// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], i64 32)
// CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 3
// CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP7]], <vscale x 16 x i8> [[TMP8]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP9]]
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z18test_svld4_vnum_u8u10__SVBool_tPKhl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP0]])
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP2]], i64 0)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], i64 16)
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], i64 32)
// CPP-CHECK-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 3
// CPP-CHECK-NEXT: [[TMP9:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP7]], <vscale x 16 x i8> [[TMP8]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP9]]
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]]
//
svuint8x4_t test_svld4_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum) MODE_ATTR
{
@ -557,30 +301,14 @@ svuint8x4_t test_svld4_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum) M
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP10]]
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld4_vnum_u16u10__SVBool_tPKtl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP8]], <vscale x 8 x i16> [[TMP9]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP10]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
svuint16x4_t test_svld4_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum) MODE_ATTR
{
@ -592,30 +320,14 @@ svuint16x4_t test_svld4_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP10]]
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld4_vnum_u32u10__SVBool_tPKjl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP8]], <vscale x 4 x i32> [[TMP9]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP10]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
svuint32x4_t test_svld4_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum) MODE_ATTR
{
@ -627,30 +339,14 @@ svuint32x4_t test_svld4_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP10]]
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld4_vnum_u64u10__SVBool_tPKml(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP8]], <vscale x 2 x i64> [[TMP9]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP10]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
svuint64x4_t test_svld4_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum) MODE_ATTR
{
@ -662,30 +358,14 @@ svuint64x4_t test_svld4_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 8)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 16)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP8]], <vscale x 8 x half> [[TMP9]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP10]]
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld4_vnum_f16u10__SVBool_tPKDhl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], i64 8)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], i64 16)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP8]], <vscale x 8 x half> [[TMP9]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP10]]
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]]
//
svfloat16x4_t test_svld4_vnum_f16(svbool_t pg, const float16_t *base, int64_t vnum) MODE_ATTR
{
@ -697,30 +377,14 @@ svfloat16x4_t test_svld4_vnum_f16(svbool_t pg, const float16_t *base, int64_t vn
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 8)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 12)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP10]]
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld4_vnum_f32u10__SVBool_tPKfl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], i64 4)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], i64 8)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP8]], <vscale x 4 x float> [[TMP9]], i64 12)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP10]]
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]]
//
svfloat32x4_t test_svld4_vnum_f32(svbool_t pg, const float32_t *base, int64_t vnum) MODE_ATTR
{
@ -732,30 +396,14 @@ svfloat32x4_t test_svld4_vnum_f32(svbool_t pg, const float32_t *base, int64_t vn
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP3]], i64 0)
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 2)
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 2
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 4)
// CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 3
// CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP8]], <vscale x 2 x double> [[TMP9]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x double> [[TMP10]]
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z19test_svld4_vnum_f64u10__SVBool_tPKdl(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP1:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP1]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP3]], i64 0)
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], i64 2)
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], i64 4)
// CPP-CHECK-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 3
// CPP-CHECK-NEXT: [[TMP10:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP8]], <vscale x 2 x double> [[TMP9]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x double> [[TMP10]]
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]]
//
svfloat64x4_t test_svld4_vnum_f64(svbool_t pg, const float64_t *base, int64_t vnum) MODE_ATTR
{

2046
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_reinterpret-bfloat.c
View File

File diff suppressed because it is too large Load Diff

10437
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_reinterpret.c
View File

File diff suppressed because it is too large Load Diff

24
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set2-bfloat.c
View File

@ -24,13 +24,17 @@
// CHECK-LABEL: @test_svset2_bf16_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svset2_bf16_014svbfloat16x2_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
svbfloat16x2_t test_svset2_bf16_0(svbfloat16x2_t tuple, svbfloat16_t x) ATTR
{
@ -39,13 +43,17 @@ svbfloat16x2_t test_svset2_bf16_0(svbfloat16x2_t tuple, svbfloat16_t x) ATTR
// CHECK-LABEL: @test_svset2_bf16_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z18test_svset2_bf16_114svbfloat16x2_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]]
//
svbfloat16x2_t test_svset2_bf16_1(svbfloat16x2_t tuple, svbfloat16_t x) ATTR
{

132
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set2.c
View File

@ -24,13 +24,17 @@
// CHECK-LABEL: @test_svset2_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z14test_svset2_s810svint8x2_tu10__SVInt8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]]
//
svint8x2_t test_svset2_s8(svint8x2_t tuple, svint8_t x) ATTR
{
@ -39,13 +43,17 @@ svint8x2_t test_svset2_s8(svint8x2_t tuple, svint8_t x) ATTR
// CHECK-LABEL: @test_svset2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svset2_s1611svint16x2_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
svint16x2_t test_svset2_s16(svint16x2_t tuple, svint16_t x) ATTR
{
@ -54,13 +62,17 @@ svint16x2_t test_svset2_s16(svint16x2_t tuple, svint16_t x) ATTR
// CHECK-LABEL: @test_svset2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svset2_s3211svint32x2_tu11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
svint32x2_t test_svset2_s32(svint32x2_t tuple, svint32_t x) ATTR
{
@ -69,13 +81,17 @@ svint32x2_t test_svset2_s32(svint32x2_t tuple, svint32_t x) ATTR
// CHECK-LABEL: @test_svset2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svset2_s6411svint64x2_tu11__SVInt64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
svint64x2_t test_svset2_s64(svint64x2_t tuple, svint64_t x) ATTR
{
@ -84,13 +100,17 @@ svint64x2_t test_svset2_s64(svint64x2_t tuple, svint64_t x) ATTR
// CHECK-LABEL: @test_svset2_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z14test_svset2_u811svuint8x2_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]]
//
svuint8x2_t test_svset2_u8(svuint8x2_t tuple, svuint8_t x) ATTR
{
@ -99,13 +119,17 @@ svuint8x2_t test_svset2_u8(svuint8x2_t tuple, svuint8_t x) ATTR
// CHECK-LABEL: @test_svset2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 16 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svset2_u1612svuint16x2_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]]
//
svuint16x2_t test_svset2_u16(svuint16x2_t tuple, svuint16_t x) ATTR
{
@ -114,13 +138,17 @@ svuint16x2_t test_svset2_u16(svuint16x2_t tuple, svuint16_t x) ATTR
// CHECK-LABEL: @test_svset2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 8 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svset2_u3212svuint32x2_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]]
//
svuint32x2_t test_svset2_u32(svuint32x2_t tuple, svuint32_t x) ATTR
{
@ -129,13 +157,17 @@ svuint32x2_t test_svset2_u32(svuint32x2_t tuple, svuint32_t x) ATTR
// CHECK-LABEL: @test_svset2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 4 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svset2_u6412svuint64x2_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]]
//
svuint64x2_t test_svset2_u64(svuint64x2_t tuple, svuint64_t x) ATTR
{
@ -144,13 +176,17 @@ svuint64x2_t test_svset2_u64(svuint64x2_t tuple, svuint64_t x) ATTR
// CHECK-LABEL: @test_svset2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 16 x half> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svset2_f1613svfloat16x2_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 16 x half> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]]
//
svfloat16x2_t test_svset2_f16(svfloat16x2_t tuple, svfloat16_t x) ATTR
{
@ -159,13 +195,17 @@ svfloat16x2_t test_svset2_f16(svfloat16x2_t tuple, svfloat16_t x) ATTR
// CHECK-LABEL: @test_svset2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 8 x float> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svset2_f3213svfloat32x2_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 8 x float> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]]
//
svfloat32x2_t test_svset2_f32(svfloat32x2_t tuple, svfloat32_t x) ATTR
{
@ -174,13 +214,17 @@ svfloat32x2_t test_svset2_f32(svfloat32x2_t tuple, svfloat32_t x) ATTR
// CHECK-LABEL: @test_svset2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 2)
// CHECK-NEXT: ret <vscale x 4 x double> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svset2_f6413svfloat64x2_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 4 x double> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]]
//
svfloat64x2_t test_svset2_f64(svfloat64x2_t tuple, svfloat64_t x) ATTR
{

42
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set3-bfloat.c
View File

@ -24,13 +24,19 @@
// CHECK-LABEL: @test_svset3_bf16_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svset3_bf16_014svbfloat16x3_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]]
//
svbfloat16x3_t test_svset3_bf16_0(svbfloat16x3_t tuple, svbfloat16_t x) ATTR
{
@ -39,13 +45,19 @@ svbfloat16x3_t test_svset3_bf16_0(svbfloat16x3_t tuple, svbfloat16_t x) ATTR
// CHECK-LABEL: @test_svset3_bf16_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svset3_bf16_114svbfloat16x3_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]]
//
svbfloat16x3_t test_svset3_bf16_1(svbfloat16x3_t tuple, svbfloat16_t x) ATTR
{
@ -54,13 +66,19 @@ svbfloat16x3_t test_svset3_bf16_1(svbfloat16x3_t tuple, svbfloat16_t x) ATTR
// CHECK-LABEL: @test_svset3_bf16_2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[X:%.*]], 2
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z18test_svset3_bf16_214svbfloat16x3_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[X:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]]
//
svbfloat16x3_t test_svset3_bf16_2(svbfloat16x3_t tuple, svbfloat16_t x) ATTR
{

154
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set3.c
View File

@ -26,13 +26,19 @@
// several parameters, one for each member of the original struct.
// CHECK-LABEL: @test_svset3_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 48 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z14test_svset3_s810svint8x3_tu10__SVInt8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 48 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
//
svint8x3_t test_svset3_s8(svint8x3_t tuple, svint8_t x) ATTR
{
@ -41,13 +47,19 @@ svint8x3_t test_svset3_s8(svint8x3_t tuple, svint8_t x) ATTR
// CHECK-LABEL: @test_svset3_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 24 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[X:%.*]], 2
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svset3_s1611svint16x3_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 24 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[X:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]]
//
svint16x3_t test_svset3_s16(svint16x3_t tuple, svint16_t x) ATTR
{
@ -56,13 +68,19 @@ svint16x3_t test_svset3_s16(svint16x3_t tuple, svint16_t x) ATTR
// CHECK-LABEL: @test_svset3_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 12 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svset3_s3211svint32x3_tu11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 12 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]]
//
svint32x3_t test_svset3_s32(svint32x3_t tuple, svint32_t x) ATTR
{
@ -71,13 +89,19 @@ svint32x3_t test_svset3_s32(svint32x3_t tuple, svint32_t x) ATTR
// CHECK-LABEL: @test_svset3_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 2)
// CHECK-NEXT: ret <vscale x 6 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svset3_s6411svint64x3_tu11__SVInt64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 6 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]]
//
svint64x3_t test_svset3_s64(svint64x3_t tuple, svint64_t x) ATTR
{
@ -86,13 +110,19 @@ svint64x3_t test_svset3_s64(svint64x3_t tuple, svint64_t x) ATTR
// CHECK-LABEL: @test_svset3_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 32)
// CHECK-NEXT: ret <vscale x 48 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[X:%.*]], 2
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z14test_svset3_u811svuint8x3_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 32)
// CPP-CHECK-NEXT: ret <vscale x 48 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[X:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
//
svuint8x3_t test_svset3_u8(svuint8x3_t tuple, svuint8_t x) ATTR
{
@ -101,13 +131,19 @@ svuint8x3_t test_svset3_u8(svuint8x3_t tuple, svuint8_t x) ATTR
// CHECK-LABEL: @test_svset3_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 24 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svset3_u1612svuint16x3_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 24 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]]
//
svuint16x3_t test_svset3_u16(svuint16x3_t tuple, svuint16_t x) ATTR
{
@ -116,13 +152,19 @@ svuint16x3_t test_svset3_u16(svuint16x3_t tuple, svuint16_t x) ATTR
// CHECK-LABEL: @test_svset3_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 12 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svset3_u3212svuint32x3_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 12 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]]
//
svuint32x3_t test_svset3_u32(svuint32x3_t tuple, svuint32_t x) ATTR
{
@ -131,13 +173,19 @@ svuint32x3_t test_svset3_u32(svuint32x3_t tuple, svuint32_t x) ATTR
// CHECK-LABEL: @test_svset3_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 6 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[X:%.*]], 2
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svset3_u6412svuint64x3_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 6 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[X:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]]
//
svuint64x3_t test_svset3_u64(svuint64x3_t tuple, svuint64_t x) ATTR
{
@ -146,13 +194,19 @@ svuint64x3_t test_svset3_u64(svuint64x3_t tuple, svuint64_t x) ATTR
// CHECK-LABEL: @test_svset3_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 24 x half> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svset3_f1613svfloat16x3_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 24 x half> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]]
//
svfloat16x3_t test_svset3_f16(svfloat16x3_t tuple, svfloat16_t x) ATTR
{
@ -161,13 +215,19 @@ svfloat16x3_t test_svset3_f16(svfloat16x3_t tuple, svfloat16_t x) ATTR
// CHECK-LABEL: @test_svset3_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 12 x float> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svset3_f3213svfloat32x3_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 12 x float> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]]
//
svfloat32x3_t test_svset3_f32(svfloat32x3_t tuple, svfloat32_t x) ATTR
{
@ -176,13 +236,19 @@ svfloat32x3_t test_svset3_f32(svfloat32x3_t tuple, svfloat32_t x) ATTR
// CHECK-LABEL: @test_svset3_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 6 x double> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[X:%.*]], 2
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z15test_svset3_f6413svfloat64x3_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 6 x double> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[X:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]]
//
svfloat64x3_t test_svset3_f64(svfloat64x3_t tuple, svfloat64_t x) ATTR
{

64
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set4-bfloat.c
View File

@ -24,13 +24,21 @@
// CHECK-LABEL: @test_svset4_bf16_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], <vscale x 8 x bfloat> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z18test_svset4_bf16_014svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], <vscale x 8 x bfloat> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
//
svbfloat16x4_t test_svset4_bf16_0(svbfloat16x4_t tuple, svbfloat16_t x) ATTR
{
@ -39,13 +47,21 @@ svbfloat16x4_t test_svset4_bf16_0(svbfloat16x4_t tuple, svbfloat16_t x) ATTR
// CHECK-LABEL: @test_svset4_bf16_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], <vscale x 8 x bfloat> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z18test_svset4_bf16_114svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], <vscale x 8 x bfloat> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
//
svbfloat16x4_t test_svset4_bf16_1(svbfloat16x4_t tuple, svbfloat16_t x) ATTR
{
@ -54,13 +70,21 @@ svbfloat16x4_t test_svset4_bf16_1(svbfloat16x4_t tuple, svbfloat16_t x) ATTR
// CHECK-LABEL: @test_svset4_bf16_2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], <vscale x 8 x bfloat> [[X:%.*]], 2
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z18test_svset4_bf16_214svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], <vscale x 8 x bfloat> [[X:%.*]], 2
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
//
svbfloat16x4_t test_svset4_bf16_2(svbfloat16x4_t tuple, svbfloat16_t x) ATTR
{
@ -69,13 +93,21 @@ svbfloat16x4_t test_svset4_bf16_2(svbfloat16x4_t tuple, svbfloat16_t x) ATTR
// CHECK-LABEL: @test_svset4_bf16_3(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], <vscale x 8 x bfloat> [[X:%.*]], 3
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z18test_svset4_bf16_314svbfloat16x4_tu14__SVBfloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], <vscale x 8 x bfloat> [[X:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
//
svbfloat16x4_t test_svset4_bf16_3(svbfloat16x4_t tuple, svbfloat16_t x) ATTR
{

176
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set4.c
View File

@ -24,13 +24,21 @@
// CHECK-LABEL: @test_svset4_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], <vscale x 16 x i8> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z14test_svset4_s810svint8x4_tu10__SVInt8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], <vscale x 16 x i8> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]]
//
svint8x4_t test_svset4_s8(svint8x4_t tuple, svint8_t x) ATTR
{
@ -39,13 +47,21 @@ svint8x4_t test_svset4_s8(svint8x4_t tuple, svint8_t x) ATTR
// CHECK-LABEL: @test_svset4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 24)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], <vscale x 8 x i16> [[X:%.*]], 3
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_s1611svint16x4_tu11__SVInt16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 24)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], <vscale x 8 x i16> [[X:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
//
svint16x4_t test_svset4_s16(svint16x4_t tuple, svint16_t x) ATTR
{
@ -54,13 +70,21 @@ svint16x4_t test_svset4_s16(svint16x4_t tuple, svint16_t x) ATTR
// CHECK-LABEL: @test_svset4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], <vscale x 4 x i32> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_s3211svint32x4_tu11__SVInt32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], <vscale x 4 x i32> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
//
svint32x4_t test_svset4_s32(svint32x4_t tuple, svint32_t x) ATTR
{
@ -69,13 +93,21 @@ svint32x4_t test_svset4_s32(svint32x4_t tuple, svint32_t x) ATTR
// CHECK-LABEL: @test_svset4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 2)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], <vscale x 2 x i64> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_s6411svint64x4_tu11__SVInt64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 2)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], <vscale x 2 x i64> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
//
svint64x4_t test_svset4_s64(svint64x4_t tuple, svint64_t x) ATTR
{
@ -84,13 +116,21 @@ svint64x4_t test_svset4_s64(svint64x4_t tuple, svint64_t x) ATTR
// CHECK-LABEL: @test_svset4_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i8> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], <vscale x 16 x i8> [[X:%.*]], 3
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z14test_svset4_u811svuint8x4_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], <vscale x 16 x i8> [[X:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP4]]
//
svuint8x4_t test_svset4_u8(svuint8x4_t tuple, svuint8_t x) ATTR
{
@ -99,13 +139,21 @@ svuint8x4_t test_svset4_u8(svuint8x4_t tuple, svuint8_t x) ATTR
// CHECK-LABEL: @test_svset4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 32 x i16> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], <vscale x 8 x i16> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_u1612svuint16x4_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], <vscale x 8 x i16> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
//
svuint16x4_t test_svset4_u16(svuint16x4_t tuple, svuint16_t x) ATTR
{
@ -114,13 +162,21 @@ svuint16x4_t test_svset4_u16(svuint16x4_t tuple, svuint16_t x) ATTR
// CHECK-LABEL: @test_svset4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 16 x i32> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], <vscale x 4 x i32> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_u3212svuint32x4_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], <vscale x 4 x i32> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
//
svuint32x4_t test_svset4_u32(svuint32x4_t tuple, svuint32_t x) ATTR
{
@ -129,13 +185,21 @@ svuint32x4_t test_svset4_u32(svuint32x4_t tuple, svuint32_t x) ATTR
// CHECK-LABEL: @test_svset4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x i64> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], <vscale x 2 x i64> [[X:%.*]], 3
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_u6412svuint64x4_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], <vscale x 2 x i64> [[X:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
//
svuint64x4_t test_svset4_u64(svuint64x4_t tuple, svuint64_t x) ATTR
{
@ -144,13 +208,21 @@ svuint64x4_t test_svset4_u64(svuint64x4_t tuple, svuint64_t x) ATTR
// CHECK-LABEL: @test_svset4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 8)
// CHECK-NEXT: ret <vscale x 32 x half> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], <vscale x 8 x half> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_f1613svfloat16x4_tu13__SVFloat16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 8)
// CPP-CHECK-NEXT: ret <vscale x 32 x half> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], <vscale x 8 x half> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]]
//
svfloat16x4_t test_svset4_f16(svfloat16x4_t tuple, svfloat16_t x) ATTR
{
@ -159,13 +231,21 @@ svfloat16x4_t test_svset4_f16(svfloat16x4_t tuple, svfloat16_t x) ATTR
// CHECK-LABEL: @test_svset4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 4)
// CHECK-NEXT: ret <vscale x 16 x float> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], <vscale x 4 x float> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_f3213svfloat32x4_tu13__SVFloat32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 4)
// CPP-CHECK-NEXT: ret <vscale x 16 x float> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], <vscale x 4 x float> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]]
//
svfloat32x4_t test_svset4_f32(svfloat32x4_t tuple, svfloat32_t x) ATTR
{
@ -174,13 +254,21 @@ svfloat32x4_t test_svset4_f32(svfloat32x4_t tuple, svfloat32_t x) ATTR
// CHECK-LABEL: @test_svset4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 6)
// CHECK-NEXT: ret <vscale x 8 x double> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], <vscale x 2 x double> [[X:%.*]], 3
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_f6413svfloat64x4_tu13__SVFloat64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 6)
// CPP-CHECK-NEXT: ret <vscale x 8 x double> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], <vscale x 2 x double> [[X:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]]
//
svfloat64x4_t test_svset4_f64(svfloat64x4_t tuple, svfloat64_t x) ATTR
{

44
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st2-bfloat.c
View File

@ -24,18 +24,22 @@
#endif
// CHECK-LABEL: @test_svst2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svst2_bf16u10__SVBool_tPu6__bf1614svbfloat16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_bf16(svbool_t pg, bfloat16_t *base, svbfloat16x2_t data) MODE_ATTR
@ -45,20 +49,24 @@ void test_svst2_bf16(svbool_t pg, bfloat16_t *base, svbfloat16x2_t data) MODE_AT
// CHECK-LABEL: @test_svst2_vnum_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svst2_vnum_bf16u10__SVBool_tPu6__bf16l14svbfloat16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_bf16(svbool_t pg, bfloat16_t *base, int64_t vnum, svbfloat16x2_t data) MODE_ATTR

468
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st2.c
View File

@ -24,16 +24,20 @@
// CHECK-LABEL: @test_svst2_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z13test_svst2_s8u10__SVBool_tPa10svint8x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_s8(svbool_t pg, int8_t *base, svint8x2_t data) MODE_ATTR
@ -43,18 +47,22 @@ void test_svst2_s8(svbool_t pg, int8_t *base, svint8x2_t data) MODE_ATTR
// CHECK-LABEL: @test_svst2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst2_s16u10__SVBool_tPs11svint16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_s16(svbool_t pg, int16_t *base, svint16x2_t data) MODE_ATTR
@ -64,18 +72,22 @@ void test_svst2_s16(svbool_t pg, int16_t *base, svint16x2_t data) MODE_ATTR
// CHECK-LABEL: @test_svst2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst2_s32u10__SVBool_tPi11svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_s32(svbool_t pg, int32_t *base, svint32x2_t data) MODE_ATTR
@ -85,18 +97,22 @@ void test_svst2_s32(svbool_t pg, int32_t *base, svint32x2_t data) MODE_ATTR
// CHECK-LABEL: @test_svst2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst2_s64u10__SVBool_tPl11svint64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_s64(svbool_t pg, int64_t *base, svint64x2_t data) MODE_ATTR
@ -106,16 +122,20 @@ void test_svst2_s64(svbool_t pg, int64_t *base, svint64x2_t data) MODE_ATTR
// CHECK-LABEL: @test_svst2_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z13test_svst2_u8u10__SVBool_tPh11svuint8x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_u8(svbool_t pg, uint8_t *base, svuint8x2_t data) MODE_ATTR
@ -125,18 +145,22 @@ void test_svst2_u8(svbool_t pg, uint8_t *base, svuint8x2_t data) MODE_ATTR
// CHECK-LABEL: @test_svst2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst2_u16u10__SVBool_tPt12svuint16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_u16(svbool_t pg, uint16_t *base, svuint16x2_t data) MODE_ATTR
@ -146,18 +170,22 @@ void test_svst2_u16(svbool_t pg, uint16_t *base, svuint16x2_t data) MODE_ATTR
// CHECK-LABEL: @test_svst2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst2_u32u10__SVBool_tPj12svuint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_u32(svbool_t pg, uint32_t *base, svuint32x2_t data) MODE_ATTR
@ -167,18 +195,22 @@ void test_svst2_u32(svbool_t pg, uint32_t *base, svuint32x2_t data) MODE_ATTR
// CHECK-LABEL: @test_svst2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst2_u64u10__SVBool_tPm12svuint64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_u64(svbool_t pg, uint64_t *base, svuint64x2_t data) MODE_ATTR
@ -188,18 +220,22 @@ void test_svst2_u64(svbool_t pg, uint64_t *base, svuint64x2_t data) MODE_ATTR
// CHECK-LABEL: @test_svst2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst2_f16u10__SVBool_tPDh13svfloat16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_f16(svbool_t pg, float16_t *base, svfloat16x2_t data) MODE_ATTR
@ -209,18 +245,22 @@ void test_svst2_f16(svbool_t pg, float16_t *base, svfloat16x2_t data) MODE_ATTR
// CHECK-LABEL: @test_svst2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst2_f32u10__SVBool_tPf13svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_f32(svbool_t pg, float32_t *base, svfloat32x2_t data) MODE_ATTR
@ -230,18 +270,22 @@ void test_svst2_f32(svbool_t pg, float32_t *base, svfloat32x2_t data) MODE_ATTR
// CHECK-LABEL: @test_svst2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst2_f64u10__SVBool_tPd13svfloat64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_f64(svbool_t pg, float64_t *base, svfloat64x2_t data) MODE_ATTR
@ -251,18 +295,22 @@ void test_svst2_f64(svbool_t pg, float64_t *base, svfloat64x2_t data) MODE_ATTR
// CHECK-LABEL: @test_svst2_vnum_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP4]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z18test_svst2_vnum_s8u10__SVBool_tPal10svint8x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP4]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_s8(svbool_t pg, int8_t *base, int64_t vnum, svint8x2_t data) MODE_ATTR
@ -272,20 +320,24 @@ void test_svst2_vnum_s8(svbool_t pg, int8_t *base, int64_t vnum, svint8x2_t data
// CHECK-LABEL: @test_svst2_vnum_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst2_vnum_s16u10__SVBool_tPsl11svint16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_s16(svbool_t pg, int16_t *base, int64_t vnum, svint16x2_t data) MODE_ATTR
@ -295,20 +347,24 @@ void test_svst2_vnum_s16(svbool_t pg, int16_t *base, int64_t vnum, svint16x2_t d
// CHECK-LABEL: @test_svst2_vnum_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst2_vnum_s32u10__SVBool_tPil11svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_s32(svbool_t pg, int32_t *base, int64_t vnum, svint32x2_t data) MODE_ATTR
@ -318,20 +374,24 @@ void test_svst2_vnum_s32(svbool_t pg, int32_t *base, int64_t vnum, svint32x2_t d
// CHECK-LABEL: @test_svst2_vnum_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst2_vnum_s64u10__SVBool_tPll11svint64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_s64(svbool_t pg, int64_t *base, int64_t vnum, svint64x2_t data) MODE_ATTR
@ -341,18 +401,22 @@ void test_svst2_vnum_s64(svbool_t pg, int64_t *base, int64_t vnum, svint64x2_t d
// CHECK-LABEL: @test_svst2_vnum_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP4]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z18test_svst2_vnum_u8u10__SVBool_tPhl11svuint8x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP4]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_u8(svbool_t pg, uint8_t *base, int64_t vnum, svuint8x2_t data) MODE_ATTR
@ -362,20 +426,24 @@ void test_svst2_vnum_u8(svbool_t pg, uint8_t *base, int64_t vnum, svuint8x2_t da
// CHECK-LABEL: @test_svst2_vnum_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst2_vnum_u16u10__SVBool_tPtl12svuint16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_u16(svbool_t pg, uint16_t *base, int64_t vnum, svuint16x2_t data) MODE_ATTR
@ -385,20 +453,24 @@ void test_svst2_vnum_u16(svbool_t pg, uint16_t *base, int64_t vnum, svuint16x2_t
// CHECK-LABEL: @test_svst2_vnum_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst2_vnum_u32u10__SVBool_tPjl12svuint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_u32(svbool_t pg, uint32_t *base, int64_t vnum, svuint32x2_t data) MODE_ATTR
@ -408,20 +480,24 @@ void test_svst2_vnum_u32(svbool_t pg, uint32_t *base, int64_t vnum, svuint32x2_t
// CHECK-LABEL: @test_svst2_vnum_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst2_vnum_u64u10__SVBool_tPml12svuint64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_u64(svbool_t pg, uint64_t *base, int64_t vnum, svuint64x2_t data) MODE_ATTR
@ -431,20 +507,24 @@ void test_svst2_vnum_u64(svbool_t pg, uint64_t *base, int64_t vnum, svuint64x2_t
// CHECK-LABEL: @test_svst2_vnum_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst2_vnum_f16u10__SVBool_tPDhl13svfloat16x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x i1> [[TMP2]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_f16(svbool_t pg, float16_t *base, int64_t vnum, svfloat16x2_t data) MODE_ATTR
@ -454,20 +534,24 @@ void test_svst2_vnum_f16(svbool_t pg, float16_t *base, int64_t vnum, svfloat16x2
// CHECK-LABEL: @test_svst2_vnum_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst2_vnum_f32u10__SVBool_tPfl13svfloat32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x i1> [[TMP2]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_f32(svbool_t pg, float32_t *base, int64_t vnum, svfloat32x2_t data) MODE_ATTR
@ -477,20 +561,24 @@ void test_svst2_vnum_f32(svbool_t pg, float32_t *base, int64_t vnum, svfloat32x2
// CHECK-LABEL: @test_svst2_vnum_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst2_vnum_f64u10__SVBool_tPdl13svfloat64x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x i1> [[TMP2]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: ret void
//
void test_svst2_vnum_f64(svbool_t pg, float64_t *base, int64_t vnum, svfloat64x2_t data) MODE_ATTR

56
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st3-bfloat.c
View File

@ -25,20 +25,26 @@
// CHECK-LABEL: @test_svst3_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svst3_bf16u10__SVBool_tPu6__bf1614svbfloat16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_bf16(svbool_t pg, bfloat16_t *base, svbfloat16x3_t data) MODE_ATTR
@ -48,22 +54,28 @@ void test_svst3_bf16(svbool_t pg, bfloat16_t *base, svbfloat16x3_t data) MODE_AT
// CHECK-LABEL: @test_svst3_vnum_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[TMP7]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svst3_vnum_bf16u10__SVBool_tPu6__bf16l14svbfloat16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[TMP7]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_bf16(svbool_t pg, bfloat16_t *base, int64_t vnum, svbfloat16x3_t data) MODE_ATTR

600
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st3.c
View File

@ -24,18 +24,24 @@
// CHECK-LABEL: @test_svst3_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z13test_svst3_s8u10__SVBool_tPa10svint8x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_s8(svbool_t pg, int8_t *base, svint8x3_t data) MODE_ATTR
@ -45,20 +51,26 @@ void test_svst3_s8(svbool_t pg, int8_t *base, svint8x3_t data) MODE_ATTR
// CHECK-LABEL: @test_svst3_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst3_s16u10__SVBool_tPs11svint16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_s16(svbool_t pg, int16_t *base, svint16x3_t data) MODE_ATTR
@ -68,20 +80,26 @@ void test_svst3_s16(svbool_t pg, int16_t *base, svint16x3_t data) MODE_ATTR
// CHECK-LABEL: @test_svst3_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst3_s32u10__SVBool_tPi11svint32x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_s32(svbool_t pg, int32_t *base, svint32x3_t data) MODE_ATTR
@ -91,20 +109,26 @@ void test_svst3_s32(svbool_t pg, int32_t *base, svint32x3_t data) MODE_ATTR
// CHECK-LABEL: @test_svst3_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst3_s64u10__SVBool_tPl11svint64x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_s64(svbool_t pg, int64_t *base, svint64x3_t data) MODE_ATTR
@ -114,18 +138,24 @@ void test_svst3_s64(svbool_t pg, int64_t *base, svint64x3_t data) MODE_ATTR
// CHECK-LABEL: @test_svst3_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z13test_svst3_u8u10__SVBool_tPh11svuint8x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_u8(svbool_t pg, uint8_t *base, svuint8x3_t data) MODE_ATTR
@ -135,20 +165,26 @@ void test_svst3_u8(svbool_t pg, uint8_t *base, svuint8x3_t data) MODE_ATTR
// CHECK-LABEL: @test_svst3_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst3_u16u10__SVBool_tPt12svuint16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_u16(svbool_t pg, uint16_t *base, svuint16x3_t data) MODE_ATTR
@ -158,20 +194,26 @@ void test_svst3_u16(svbool_t pg, uint16_t *base, svuint16x3_t data) MODE_ATTR
// CHECK-LABEL: @test_svst3_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst3_u32u10__SVBool_tPj12svuint32x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_u32(svbool_t pg, uint32_t *base, svuint32x3_t data) MODE_ATTR
@ -181,20 +223,26 @@ void test_svst3_u32(svbool_t pg, uint32_t *base, svuint32x3_t data) MODE_ATTR
// CHECK-LABEL: @test_svst3_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst3_u64u10__SVBool_tPm12svuint64x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_u64(svbool_t pg, uint64_t *base, svuint64x3_t data) MODE_ATTR
@ -204,20 +252,26 @@ void test_svst3_u64(svbool_t pg, uint64_t *base, svuint64x3_t data) MODE_ATTR
// CHECK-LABEL: @test_svst3_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst3_f16u10__SVBool_tPDh13svfloat16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_f16(svbool_t pg, float16_t *base, svfloat16x3_t data) MODE_ATTR
@ -227,20 +281,26 @@ void test_svst3_f16(svbool_t pg, float16_t *base, svfloat16x3_t data) MODE_ATTR
// CHECK-LABEL: @test_svst3_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst3_f32u10__SVBool_tPf13svfloat32x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_f32(svbool_t pg, float32_t *base, svfloat32x3_t data) MODE_ATTR
@ -250,20 +310,26 @@ void test_svst3_f32(svbool_t pg, float32_t *base, svfloat32x3_t data) MODE_ATTR
// CHECK-LABEL: @test_svst3_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst3_f64u10__SVBool_tPd13svfloat64x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_f64(svbool_t pg, float64_t *base, svfloat64x3_t data) MODE_ATTR
@ -273,20 +339,26 @@ void test_svst3_f64(svbool_t pg, float64_t *base, svfloat64x3_t data) MODE_ATTR
// CHECK-LABEL: @test_svst3_vnum_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP6]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z18test_svst3_vnum_s8u10__SVBool_tPal10svint8x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP6]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_s8(svbool_t pg, int8_t *base, int64_t vnum, svint8x3_t data) MODE_ATTR
@ -296,22 +368,28 @@ void test_svst3_vnum_s8(svbool_t pg, int8_t *base, int64_t vnum, svint8x3_t data
// CHECK-LABEL: @test_svst3_vnum_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[TMP7]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst3_vnum_s16u10__SVBool_tPsl11svint16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[TMP7]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_s16(svbool_t pg, int16_t *base, int64_t vnum, svint16x3_t data) MODE_ATTR
@ -321,22 +399,28 @@ void test_svst3_vnum_s16(svbool_t pg, int16_t *base, int64_t vnum, svint16x3_t d
// CHECK-LABEL: @test_svst3_vnum_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[TMP7]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst3_vnum_s32u10__SVBool_tPil11svint32x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[TMP7]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_s32(svbool_t pg, int32_t *base, int64_t vnum, svint32x3_t data) MODE_ATTR
@ -346,22 +430,28 @@ void test_svst3_vnum_s32(svbool_t pg, int32_t *base, int64_t vnum, svint32x3_t d
// CHECK-LABEL: @test_svst3_vnum_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[TMP7]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst3_vnum_s64u10__SVBool_tPll11svint64x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[TMP7]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_s64(svbool_t pg, int64_t *base, int64_t vnum, svint64x3_t data) MODE_ATTR
@ -371,20 +461,26 @@ void test_svst3_vnum_s64(svbool_t pg, int64_t *base, int64_t vnum, svint64x3_t d
// CHECK-LABEL: @test_svst3_vnum_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP3]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP6]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z18test_svst3_vnum_u8u10__SVBool_tPhl11svuint8x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP3]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP6]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_u8(svbool_t pg, uint8_t *base, int64_t vnum, svuint8x3_t data) MODE_ATTR
@ -394,22 +490,28 @@ void test_svst3_vnum_u8(svbool_t pg, uint8_t *base, int64_t vnum, svuint8x3_t da
// CHECK-LABEL: @test_svst3_vnum_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[TMP7]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst3_vnum_u16u10__SVBool_tPtl12svuint16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[TMP7]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_u16(svbool_t pg, uint16_t *base, int64_t vnum, svuint16x3_t data) MODE_ATTR
@ -419,22 +521,28 @@ void test_svst3_vnum_u16(svbool_t pg, uint16_t *base, int64_t vnum, svuint16x3_t
// CHECK-LABEL: @test_svst3_vnum_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[TMP7]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst3_vnum_u32u10__SVBool_tPjl12svuint32x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[TMP7]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_u32(svbool_t pg, uint32_t *base, int64_t vnum, svuint32x3_t data) MODE_ATTR
@ -444,22 +552,28 @@ void test_svst3_vnum_u32(svbool_t pg, uint32_t *base, int64_t vnum, svuint32x3_t
// CHECK-LABEL: @test_svst3_vnum_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[TMP7]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst3_vnum_u64u10__SVBool_tPml12svuint64x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[TMP7]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_u64(svbool_t pg, uint64_t *base, int64_t vnum, svuint64x3_t data) MODE_ATTR
@ -469,22 +583,28 @@ void test_svst3_vnum_u64(svbool_t pg, uint64_t *base, int64_t vnum, svuint64x3_t
// CHECK-LABEL: @test_svst3_vnum_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[TMP7]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst3_vnum_f16u10__SVBool_tPDhl13svfloat16x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x i1> [[TMP3]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x i1> [[TMP6]], ptr [[TMP7]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_f16(svbool_t pg, float16_t *base, int64_t vnum, svfloat16x3_t data) MODE_ATTR
@ -494,22 +614,28 @@ void test_svst3_vnum_f16(svbool_t pg, float16_t *base, int64_t vnum, svfloat16x3
// CHECK-LABEL: @test_svst3_vnum_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[TMP7]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst3_vnum_f32u10__SVBool_tPfl13svfloat32x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x i1> [[TMP3]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x i1> [[TMP6]], ptr [[TMP7]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_f32(svbool_t pg, float32_t *base, int64_t vnum, svfloat32x3_t data) MODE_ATTR
@ -519,22 +645,28 @@ void test_svst3_vnum_f32(svbool_t pg, float32_t *base, int64_t vnum, svfloat32x3
// CHECK-LABEL: @test_svst3_vnum_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 2
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[TMP7]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst3_vnum_f64u10__SVBool_tPdl13svfloat64x3_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x i1> [[TMP3]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 0
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], 2
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP7:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x i1> [[TMP6]], ptr [[TMP7]])
// CPP-CHECK-NEXT: ret void
//
void test_svst3_vnum_f64(svbool_t pg, float64_t *base, int64_t vnum, svfloat64x3_t data) MODE_ATTR

68
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st4-bfloat.c
View File

@ -25,22 +25,30 @@
// CHECK-LABEL: @test_svst4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z15test_svst4_bf16u10__SVBool_tPu6__bf1614svbfloat16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_bf16(svbool_t pg, bfloat16_t *base, svbfloat16x4_t data) MODE_ATTR
@ -50,24 +58,32 @@ void test_svst4_bf16(svbool_t pg, bfloat16_t *base, svbfloat16x4_t data) MODE_AT
// CHECK-LABEL: @test_svst4_vnum_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[TMP9]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z20test_svst4_vnum_bf16u10__SVBool_tPu6__bf16l14svbfloat16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } poison, <vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP0]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], <vscale x 8 x bfloat> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP2]], <vscale x 8 x bfloat> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 8 x bfloat>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x bfloat> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[TMP9]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_bf16(svbool_t pg, bfloat16_t *base, int64_t vnum, svbfloat16x4_t data) MODE_ATTR

732
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st4.c
View File

@ -24,20 +24,28 @@
// CHECK-LABEL: @test_svst4_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 3
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z13test_svst4_s8u10__SVBool_tPa10svint8x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 3
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_s8(svbool_t pg, int8_t *base, svint8x4_t data) MODE_ATTR
@ -47,22 +55,30 @@ void test_svst4_s8(svbool_t pg, int8_t *base, svint8x4_t data) MODE_ATTR
// CHECK-LABEL: @test_svst4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst4_s16u10__SVBool_tPs11svint16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_s16(svbool_t pg, int16_t *base, svint16x4_t data) MODE_ATTR
@ -72,22 +88,30 @@ void test_svst4_s16(svbool_t pg, int16_t *base, svint16x4_t data) MODE_ATTR
// CHECK-LABEL: @test_svst4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst4_s32u10__SVBool_tPi11svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_s32(svbool_t pg, int32_t *base, svint32x4_t data) MODE_ATTR
@ -97,22 +121,30 @@ void test_svst4_s32(svbool_t pg, int32_t *base, svint32x4_t data) MODE_ATTR
// CHECK-LABEL: @test_svst4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst4_s64u10__SVBool_tPl11svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_s64(svbool_t pg, int64_t *base, svint64x4_t data) MODE_ATTR
@ -122,20 +154,28 @@ void test_svst4_s64(svbool_t pg, int64_t *base, svint64x4_t data) MODE_ATTR
// CHECK-LABEL: @test_svst4_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 3
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z13test_svst4_u8u10__SVBool_tPh11svuint8x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 3
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], <vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_u8(svbool_t pg, uint8_t *base, svuint8x4_t data) MODE_ATTR
@ -145,22 +185,30 @@ void test_svst4_u8(svbool_t pg, uint8_t *base, svuint8x4_t data) MODE_ATTR
// CHECK-LABEL: @test_svst4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst4_u16u10__SVBool_tPt12svuint16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_u16(svbool_t pg, uint16_t *base, svuint16x4_t data) MODE_ATTR
@ -170,22 +218,30 @@ void test_svst4_u16(svbool_t pg, uint16_t *base, svuint16x4_t data) MODE_ATTR
// CHECK-LABEL: @test_svst4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst4_u32u10__SVBool_tPj12svuint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_u32(svbool_t pg, uint32_t *base, svuint32x4_t data) MODE_ATTR
@ -195,22 +251,30 @@ void test_svst4_u32(svbool_t pg, uint32_t *base, svuint32x4_t data) MODE_ATTR
// CHECK-LABEL: @test_svst4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst4_u64u10__SVBool_tPm12svuint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_u64(svbool_t pg, uint64_t *base, svuint64x4_t data) MODE_ATTR
@ -220,22 +284,30 @@ void test_svst4_u64(svbool_t pg, uint64_t *base, svuint64x4_t data) MODE_ATTR
// CHECK-LABEL: @test_svst4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst4_f16u10__SVBool_tPDh13svfloat16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_f16(svbool_t pg, float16_t *base, svfloat16x4_t data) MODE_ATTR
@ -245,22 +317,30 @@ void test_svst4_f16(svbool_t pg, float16_t *base, svfloat16x4_t data) MODE_ATTR
// CHECK-LABEL: @test_svst4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst4_f32u10__SVBool_tPf13svfloat32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_f32(svbool_t pg, float32_t *base, svfloat32x4_t data) MODE_ATTR
@ -270,22 +350,30 @@ void test_svst4_f32(svbool_t pg, float32_t *base, svfloat32x4_t data) MODE_ATTR
// CHECK-LABEL: @test_svst4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z14test_svst4_f64u10__SVBool_tPd13svfloat64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[BASE:%.*]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_f64(svbool_t pg, float64_t *base, svfloat64x4_t data) MODE_ATTR
@ -295,22 +383,30 @@ void test_svst4_f64(svbool_t pg, float64_t *base, svfloat64x4_t data) MODE_ATTR
// CHECK-LABEL: @test_svst4_vnum_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP8]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z18test_svst4_vnum_s8u10__SVBool_tPal10svint8x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP8]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_s8(svbool_t pg, int8_t *base, int64_t vnum, svint8x4_t data) MODE_ATTR
@ -320,24 +416,32 @@ void test_svst4_vnum_s8(svbool_t pg, int8_t *base, int64_t vnum, svint8x4_t data
// CHECK-LABEL: @test_svst4_vnum_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[TMP9]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst4_vnum_s16u10__SVBool_tPsl11svint16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[TMP9]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_s16(svbool_t pg, int16_t *base, int64_t vnum, svint16x4_t data) MODE_ATTR
@ -347,24 +451,32 @@ void test_svst4_vnum_s16(svbool_t pg, int16_t *base, int64_t vnum, svint16x4_t d
// CHECK-LABEL: @test_svst4_vnum_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[TMP9]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst4_vnum_s32u10__SVBool_tPil11svint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[TMP9]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_s32(svbool_t pg, int32_t *base, int64_t vnum, svint32x4_t data) MODE_ATTR
@ -374,24 +486,32 @@ void test_svst4_vnum_s32(svbool_t pg, int32_t *base, int64_t vnum, svint32x4_t d
// CHECK-LABEL: @test_svst4_vnum_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[TMP9]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst4_vnum_s64u10__SVBool_tPll11svint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[TMP9]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_s64(svbool_t pg, int64_t *base, int64_t vnum, svint64x4_t data) MODE_ATTR
@ -401,22 +521,30 @@ void test_svst4_vnum_s64(svbool_t pg, int64_t *base, int64_t vnum, svint64x4_t d
// CHECK-LABEL: @test_svst4_vnum_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
// CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP4]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP8]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z18test_svst4_vnum_u8u10__SVBool_tPhl11svuint8x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP4]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } poison, <vscale x 16 x i8> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP1]], <vscale x 16 x i8> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], <vscale x 16 x i8> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = getelementptr <vscale x 16 x i8>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], <vscale x 16 x i1> [[PG:%.*]], ptr [[TMP8]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_u8(svbool_t pg, uint8_t *base, int64_t vnum, svuint8x4_t data) MODE_ATTR
@ -426,24 +554,32 @@ void test_svst4_vnum_u8(svbool_t pg, uint8_t *base, int64_t vnum, svuint8x4_t da
// CHECK-LABEL: @test_svst4_vnum_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[TMP9]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst4_vnum_u16u10__SVBool_tPtl12svuint16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } poison, <vscale x 8 x i16> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP0]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], <vscale x 8 x i16> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP2]], <vscale x 8 x i16> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 8 x i16>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i16> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[TMP9]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_u16(svbool_t pg, uint16_t *base, int64_t vnum, svuint16x4_t data) MODE_ATTR
@ -453,24 +589,32 @@ void test_svst4_vnum_u16(svbool_t pg, uint16_t *base, int64_t vnum, svuint16x4_t
// CHECK-LABEL: @test_svst4_vnum_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[TMP9]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst4_vnum_u32u10__SVBool_tPjl12svuint32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } poison, <vscale x 4 x i32> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP0]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], <vscale x 4 x i32> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP2]], <vscale x 4 x i32> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 4 x i32>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i32> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[TMP9]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_u32(svbool_t pg, uint32_t *base, int64_t vnum, svuint32x4_t data) MODE_ATTR
@ -480,24 +624,32 @@ void test_svst4_vnum_u32(svbool_t pg, uint32_t *base, int64_t vnum, svuint32x4_t
// CHECK-LABEL: @test_svst4_vnum_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[TMP9]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst4_vnum_u64u10__SVBool_tPml12svuint64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } poison, <vscale x 2 x i64> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP0]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], <vscale x 2 x i64> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP2]], <vscale x 2 x i64> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 2 x i64>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i64> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[TMP9]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_u64(svbool_t pg, uint64_t *base, int64_t vnum, svuint64x4_t data) MODE_ATTR
@ -507,24 +659,32 @@ void test_svst4_vnum_u64(svbool_t pg, uint64_t *base, int64_t vnum, svuint64x4_t
// CHECK-LABEL: @test_svst4_vnum_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 24)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[TMP9]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst4_vnum_f16u10__SVBool_tPDhl13svfloat16x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 24)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } poison, <vscale x 8 x half> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP0]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], <vscale x 8 x half> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP2]], <vscale x 8 x half> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 8 x half>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]], <vscale x 8 x i1> [[TMP8]], ptr [[TMP9]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_f16(svbool_t pg, float16_t *base, int64_t vnum, svfloat16x4_t data) MODE_ATTR
@ -534,24 +694,32 @@ void test_svst4_vnum_f16(svbool_t pg, float16_t *base, int64_t vnum, svfloat16x4
// CHECK-LABEL: @test_svst4_vnum_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 12)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[TMP9]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst4_vnum_f32u10__SVBool_tPfl13svfloat32x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 12)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } poison, <vscale x 4 x float> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP0]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], <vscale x 4 x float> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP2]], <vscale x 4 x float> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 4 x float>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x float> [[TMP7]], <vscale x 4 x i1> [[TMP8]], ptr [[TMP9]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_f32(svbool_t pg, float32_t *base, int64_t vnum, svfloat32x4_t data) MODE_ATTR
@ -561,24 +729,32 @@ void test_svst4_vnum_f32(svbool_t pg, float32_t *base, int64_t vnum, svfloat32x4
// CHECK-LABEL: @test_svst4_vnum_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 6)
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[DATA_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[DATA_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 0
// CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 1
// CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 2
// CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 3
// CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[TMP9]])
// CHECK-NEXT: ret void
//
// CPP-CHECK-LABEL: @_Z19test_svst4_vnum_f64u10__SVBool_tPdl13svfloat64x4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 6)
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP5:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x i1> [[TMP4]], ptr [[TMP5]])
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } poison, <vscale x 2 x double> [[DATA_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP0]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], <vscale x 2 x double> [[DATA_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP2]], <vscale x 2 x double> [[DATA_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 0
// CPP-CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 1
// CPP-CHECK-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 2
// CPP-CHECK-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 3
// CPP-CHECK-NEXT: [[TMP8:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
// CPP-CHECK-NEXT: [[TMP9:%.*]] = getelementptr <vscale x 2 x double>, ptr [[BASE:%.*]], i64 [[VNUM:%.*]]
// CPP-CHECK-NEXT: tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x double> [[TMP7]], <vscale x 2 x i1> [[TMP8]], ptr [[TMP9]])
// CPP-CHECK-NEXT: ret void
//
void test_svst4_vnum_f64(svbool_t pg, float64_t *base, int64_t vnum, svfloat64x4_t data) MODE_ATTR

4
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_undef2-bfloat.c
View File

@ -16,11 +16,11 @@
// CHECK-LABEL: @test_svundef2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 16 x bfloat> undef
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } undef
//
// CPP-CHECK-LABEL: @_Z18test_svundef2_bf16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 16 x bfloat> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } undef
//
svbfloat16x2_t test_svundef2_bf16(void) MODE_ATTR
{

44
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_undef2.c
View File

@ -16,11 +16,11 @@
// CHECK-LABEL: @test_svundef2_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 32 x i8> undef
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
// CPP-CHECK-LABEL: @_Z16test_svundef2_s8v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> undef
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
svint8x2_t test_svundef2_s8(void) MODE_ATTR
{
@ -29,11 +29,11 @@ svint8x2_t test_svundef2_s8(void) MODE_ATTR
// CHECK-LABEL: @test_svundef2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 16 x i16> undef
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef2_s16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
svint16x2_t test_svundef2_s16(void) MODE_ATTR
{
@ -42,11 +42,11 @@ svint16x2_t test_svundef2_s16(void) MODE_ATTR
// CHECK-LABEL: @test_svundef2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 8 x i32> undef
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef2_s32v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> undef
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
svint32x2_t test_svundef2_s32(void) MODE_ATTR
{
@ -55,11 +55,11 @@ svint32x2_t test_svundef2_s32(void) MODE_ATTR
// CHECK-LABEL: @test_svundef2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 4 x i64> undef
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef2_s64v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> undef
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
svint64x2_t test_svundef2_s64(void) MODE_ATTR
{
@ -68,11 +68,11 @@ svint64x2_t test_svundef2_s64(void) MODE_ATTR
// CHECK-LABEL: @test_svundef2_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 32 x i8> undef
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
// CPP-CHECK-LABEL: @_Z16test_svundef2_u8v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 32 x i8> undef
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
svuint8x2_t test_svundef2_u8(void) MODE_ATTR
{
@ -81,11 +81,11 @@ svuint8x2_t test_svundef2_u8(void) MODE_ATTR
// CHECK-LABEL: @test_svundef2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 16 x i16> undef
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef2_u16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 16 x i16> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
svuint16x2_t test_svundef2_u16(void) MODE_ATTR
{
@ -94,11 +94,11 @@ svuint16x2_t test_svundef2_u16(void) MODE_ATTR
// CHECK-LABEL: @test_svundef2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 8 x i32> undef
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef2_u32v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 8 x i32> undef
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
svuint32x2_t test_svundef2_u32(void) MODE_ATTR
{
@ -107,11 +107,11 @@ svuint32x2_t test_svundef2_u32(void) MODE_ATTR
// CHECK-LABEL: @test_svundef2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 4 x i64> undef
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef2_u64v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 4 x i64> undef
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
svuint64x2_t test_svundef2_u64(void) MODE_ATTR
{
@ -120,11 +120,11 @@ svuint64x2_t test_svundef2_u64(void) MODE_ATTR
// CHECK-LABEL: @test_svundef2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 16 x half> undef
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef2_f16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 16 x half> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } undef
//
svfloat16x2_t test_svundef2_f16(void) MODE_ATTR
{
@ -133,11 +133,11 @@ svfloat16x2_t test_svundef2_f16(void) MODE_ATTR
// CHECK-LABEL: @test_svundef2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 8 x float> undef
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef2_f32v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 8 x float> undef
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } undef
//
svfloat32x2_t test_svundef2_f32(void) MODE_ATTR
{
@ -146,11 +146,11 @@ svfloat32x2_t test_svundef2_f32(void) MODE_ATTR
// CHECK-LABEL: @test_svundef2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 4 x double> undef
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef2_f64v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 4 x double> undef
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } undef
//
svfloat64x2_t test_svundef2_f64(void) MODE_ATTR
{

4
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_undef3-bfloat.c
View File

@ -16,11 +16,11 @@
// CHECK-LABEL: @test_svundef3_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 24 x bfloat> undef
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } undef
//
// CPP-CHECK-LABEL: @_Z18test_svundef3_bf16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 24 x bfloat> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } undef
//
svbfloat16x3_t test_svundef3_bf16(void) MODE_ATTR
{

44
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_undef3.c
View File

@ -16,11 +16,11 @@
// CHECK-LABEL: @test_svundef3_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 48 x i8> undef
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
// CPP-CHECK-LABEL: @_Z16test_svundef3_s8v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 48 x i8> undef
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
svint8x3_t test_svundef3_s8(void) MODE_ATTR
{
@ -29,11 +29,11 @@ svint8x3_t test_svundef3_s8(void) MODE_ATTR
// CHECK-LABEL: @test_svundef3_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 24 x i16> undef
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef3_s16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 24 x i16> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
svint16x3_t test_svundef3_s16(void) MODE_ATTR
{
@ -42,11 +42,11 @@ svint16x3_t test_svundef3_s16(void) MODE_ATTR
// CHECK-LABEL: @test_svundef3_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 12 x i32> undef
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef3_s32v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 12 x i32> undef
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
svint32x3_t test_svundef3_s32(void) MODE_ATTR
{
@ -55,11 +55,11 @@ svint32x3_t test_svundef3_s32(void) MODE_ATTR
// CHECK-LABEL: @test_svundef3_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 6 x i64> undef
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef3_s64v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 6 x i64> undef
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
svint64x3_t test_svundef3_s64(void) MODE_ATTR
{
@ -68,11 +68,11 @@ svint64x3_t test_svundef3_s64(void) MODE_ATTR
// CHECK-LABEL: @test_svundef3_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 48 x i8> undef
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
// CPP-CHECK-LABEL: @_Z16test_svundef3_u8v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 48 x i8> undef
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
svuint8x3_t test_svundef3_u8(void) MODE_ATTR
{
@ -81,11 +81,11 @@ svuint8x3_t test_svundef3_u8(void) MODE_ATTR
// CHECK-LABEL: @test_svundef3_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 24 x i16> undef
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef3_u16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 24 x i16> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
svuint16x3_t test_svundef3_u16(void) MODE_ATTR
{
@ -94,11 +94,11 @@ svuint16x3_t test_svundef3_u16(void) MODE_ATTR
// CHECK-LABEL: @test_svundef3_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 12 x i32> undef
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef3_u32v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 12 x i32> undef
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
svuint32x3_t test_svundef3_u32(void) MODE_ATTR
{
@ -107,11 +107,11 @@ svuint32x3_t test_svundef3_u32(void) MODE_ATTR
// CHECK-LABEL: @test_svundef3_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 6 x i64> undef
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef3_u64v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 6 x i64> undef
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
svuint64x3_t test_svundef3_u64(void) MODE_ATTR
{
@ -120,11 +120,11 @@ svuint64x3_t test_svundef3_u64(void) MODE_ATTR
// CHECK-LABEL: @test_svundef3_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 24 x half> undef
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef3_f16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 24 x half> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } undef
//
svfloat16x3_t test_svundef3_f16(void) MODE_ATTR
{
@ -133,11 +133,11 @@ svfloat16x3_t test_svundef3_f16(void) MODE_ATTR
// CHECK-LABEL: @test_svundef3_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 12 x float> undef
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef3_f32v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 12 x float> undef
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } undef
//
svfloat32x3_t test_svundef3_f32(void) MODE_ATTR
{
@ -146,11 +146,11 @@ svfloat32x3_t test_svundef3_f32(void) MODE_ATTR
// CHECK-LABEL: @test_svundef3_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 6 x double> undef
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef3_f64v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 6 x double> undef
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } undef
//
svfloat64x3_t test_svundef3_f64(void) MODE_ATTR
{

4
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_undef4-bfloat.c
View File

@ -16,11 +16,11 @@
// CHECK-LABEL: @test_svundef4_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 32 x bfloat> undef
// CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } undef
//
// CPP-CHECK-LABEL: @_Z18test_svundef4_bf16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 32 x bfloat> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } undef
//
svbfloat16x4_t test_svundef4_bf16(void) MODE_ATTR
{

44
clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_undef4.c
View File

@ -16,11 +16,11 @@
// CHECK-LABEL: @test_svundef4_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 64 x i8> undef
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
// CPP-CHECK-LABEL: @_Z16test_svundef4_s8v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> undef
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
svint8x4_t test_svundef4_s8(void) MODE_ATTR
{
@ -29,11 +29,11 @@ svint8x4_t test_svundef4_s8(void) MODE_ATTR
// CHECK-LABEL: @test_svundef4_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 32 x i16> undef
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef4_s16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
svint16x4_t test_svundef4_s16(void) MODE_ATTR
{
@ -42,11 +42,11 @@ svint16x4_t test_svundef4_s16(void) MODE_ATTR
// CHECK-LABEL: @test_svundef4_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 16 x i32> undef
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef4_s32v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> undef
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
svint32x4_t test_svundef4_s32(void) MODE_ATTR
{
@ -55,11 +55,11 @@ svint32x4_t test_svundef4_s32(void) MODE_ATTR
// CHECK-LABEL: @test_svundef4_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 8 x i64> undef
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef4_s64v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> undef
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
svint64x4_t test_svundef4_s64(void) MODE_ATTR
{
@ -68,11 +68,11 @@ svint64x4_t test_svundef4_s64(void) MODE_ATTR
// CHECK-LABEL: @test_svundef4_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 64 x i8> undef
// CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
// CPP-CHECK-LABEL: @_Z16test_svundef4_u8v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 64 x i8> undef
// CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } undef
//
svuint8x4_t test_svundef4_u8(void) MODE_ATTR
{
@ -81,11 +81,11 @@ svuint8x4_t test_svundef4_u8(void) MODE_ATTR
// CHECK-LABEL: @test_svundef4_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 32 x i16> undef
// CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef4_u16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 32 x i16> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } undef
//
svuint16x4_t test_svundef4_u16(void) MODE_ATTR
{
@ -94,11 +94,11 @@ svuint16x4_t test_svundef4_u16(void) MODE_ATTR
// CHECK-LABEL: @test_svundef4_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 16 x i32> undef
// CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef4_u32v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 16 x i32> undef
// CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } undef
//
svuint32x4_t test_svundef4_u32(void) MODE_ATTR
{
@ -107,11 +107,11 @@ svuint32x4_t test_svundef4_u32(void) MODE_ATTR
// CHECK-LABEL: @test_svundef4_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 8 x i64> undef
// CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef4_u64v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 8 x i64> undef
// CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } undef
//
svuint64x4_t test_svundef4_u64(void) MODE_ATTR
{
@ -120,11 +120,11 @@ svuint64x4_t test_svundef4_u64(void) MODE_ATTR
// CHECK-LABEL: @test_svundef4_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 32 x half> undef
// CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef4_f16v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 32 x half> undef
// CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } undef
//
svfloat16x4_t test_svundef4_f16(void) MODE_ATTR
{
@ -133,11 +133,11 @@ svfloat16x4_t test_svundef4_f16(void) MODE_ATTR
// CHECK-LABEL: @test_svundef4_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 16 x float> undef
// CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef4_f32v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 16 x float> undef
// CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } undef
//
svfloat32x4_t test_svundef4_f32(void) MODE_ATTR
{
@ -146,11 +146,11 @@ svfloat32x4_t test_svundef4_f32(void) MODE_ATTR
// CHECK-LABEL: @test_svundef4_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: ret <vscale x 8 x double> undef
// CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } undef
//
// CPP-CHECK-LABEL: @_Z17test_svundef4_f64v(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: ret <vscale x 8 x double> undef
// CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } undef
//
svfloat64x4_t test_svundef4_f64(void) MODE_ATTR
{

48
clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_luti.c
View File

@ -248,17 +248,13 @@ svbfloat16_t test_svluti4_lane_bf16(svbfloat16_t table, svuint8_t indices) MODE_
// SME-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-LABEL: @test_svluti4_lane_s16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TABLE:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TABLE]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.luti4.lane.x2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]], i32 0)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.luti4.lane.x2.nxv8i16(<vscale x 8 x i16> [[TABLE_COERCE0:%.*]], <vscale x 8 x i16> [[TABLE_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]], i32 0)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_s16_x211svint16x2_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TABLE:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TABLE]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.luti4.lane.x2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]], i32 0)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.luti4.lane.x2.nxv8i16(<vscale x 8 x i16> [[TABLE_COERCE0:%.*]], <vscale x 8 x i16> [[TABLE_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]], i32 0)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svint16_t test_svluti4_lane_s16_x2(svint16x2_t table, svuint8_t indices) MODE_ATTR{
return SVE_ACLE_FUNC(svluti4_lane,_s16,_x2)(table, indices, 0);
@ -272,17 +268,13 @@ svint16_t test_svluti4_lane_s16_x2(svint16x2_t table, svuint8_t indices) MODE_AT
// SME-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-LABEL: @test_svluti4_lane_u16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TABLE:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TABLE]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.luti4.lane.x2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]], i32 3)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.luti4.lane.x2.nxv8i16(<vscale x 8 x i16> [[TABLE_COERCE0:%.*]], <vscale x 8 x i16> [[TABLE_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]], i32 3)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_u16_x212svuint16x2_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TABLE:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TABLE]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.luti4.lane.x2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]], i32 3)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.luti4.lane.x2.nxv8i16(<vscale x 8 x i16> [[TABLE_COERCE0:%.*]], <vscale x 8 x i16> [[TABLE_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]], i32 3)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_svluti4_lane_u16_x2(svuint16x2_t table, svuint8_t indices) MODE_ATTR{
return SVE_ACLE_FUNC(svluti4_lane,_u16,_x2)(table, indices, 3);
@ -296,17 +288,13 @@ svuint16_t test_svluti4_lane_u16_x2(svuint16x2_t table, svuint8_t indices) MODE_
// SME-CHECK-NEXT: ret <vscale x 8 x half> [[TMP2]]
// CHECK-LABEL: @test_svluti4_lane_f16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[TABLE:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[TABLE]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.luti4.lane.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]], i32 2)
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.luti4.lane.x2.nxv8f16(<vscale x 8 x half> [[TABLE_COERCE0:%.*]], <vscale x 8 x half> [[TABLE_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]], i32 2)
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_f16_x213svfloat16x2_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[TABLE:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[TABLE]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.luti4.lane.x2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]], i32 2)
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.luti4.lane.x2.nxv8f16(<vscale x 8 x half> [[TABLE_COERCE0:%.*]], <vscale x 8 x half> [[TABLE_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]], i32 2)
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
//
svfloat16_t test_svluti4_lane_f16_x2(svfloat16x2_t table, svuint8_t indices) MODE_ATTR{
return SVE_ACLE_FUNC(svluti4_lane,_f16,_x2)(table, indices, 2);
@ -320,17 +308,13 @@ svfloat16_t test_svluti4_lane_f16_x2(svfloat16x2_t table, svuint8_t indices) MOD
// SME-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
// CHECK-LABEL: @test_svluti4_lane_bf16_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TABLE:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TABLE]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.luti4.lane.x2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]], i32 1)
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.luti4.lane.x2.nxv8bf16(<vscale x 8 x bfloat> [[TABLE_COERCE0:%.*]], <vscale x 8 x bfloat> [[TABLE_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]], i32 1)
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z25test_svluti4_lane_bf16_x214svbfloat16x2_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TABLE:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TABLE]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.luti4.lane.x2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]], i32 1)
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.luti4.lane.x2.nxv8bf16(<vscale x 8 x bfloat> [[TABLE_COERCE0:%.*]], <vscale x 8 x bfloat> [[TABLE_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]], i32 1)
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
//
svbfloat16_t test_svluti4_lane_bf16_x2(svbfloat16x2_t table, svuint8_t indices) MODE_ATTR{
return SVE_ACLE_FUNC(svluti4_lane,_bf16,_x2)(table, indices, 1);

12
clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_tbl2-bfloat.c
View File

@ -16,17 +16,13 @@
// CHECK-LABEL: @test_svtbl2_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tbl2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tbl2.nxv8bf16(<vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], <vscale x 8 x i16> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z16test_svtbl2_bf1614svbfloat16x2_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tbl2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tbl2.nxv8bf16(<vscale x 8 x bfloat> [[DATA_COERCE0:%.*]], <vscale x 8 x bfloat> [[DATA_COERCE1:%.*]], <vscale x 8 x i16> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
//
svbfloat16_t test_svtbl2_bf16(svbfloat16x2_t data, svuint16_t indices) {
return SVE_ACLE_FUNC(svtbl2, _bf16, , )(data, indices);

132
clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_tbl2.c
View File

@ -17,17 +17,13 @@
// CHECK-LABEL: @test_svtbl2_s8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[DATA_COERCE0:%.*]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z14test_svtbl2_s810svint8x2_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[DATA_COERCE0:%.*]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svint8_t test_svtbl2_s8(svint8x2_t data, svuint8_t indices)
{
@ -36,17 +32,13 @@ svint8_t test_svtbl2_s8(svint8x2_t data, svuint8_t indices)
// CHECK-LABEL: @test_svtbl2_s16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[DATA_COERCE0:%.*]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], <vscale x 8 x i16> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z15test_svtbl2_s1611svint16x2_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[DATA_COERCE0:%.*]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], <vscale x 8 x i16> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svint16_t test_svtbl2_s16(svint16x2_t data, svuint16_t indices)
{
@ -55,17 +47,13 @@ svint16_t test_svtbl2_s16(svint16x2_t data, svuint16_t indices)
// CHECK-LABEL: @test_svtbl2_s32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[DATA_COERCE0:%.*]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], <vscale x 4 x i32> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z15test_svtbl2_s3211svint32x2_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[DATA_COERCE0:%.*]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], <vscale x 4 x i32> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
//
svint32_t test_svtbl2_s32(svint32x2_t data, svuint32_t indices)
{
@ -74,17 +62,13 @@ svint32_t test_svtbl2_s32(svint32x2_t data, svuint32_t indices)
// CHECK-LABEL: @test_svtbl2_s64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[DATA_COERCE0:%.*]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], <vscale x 2 x i64> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z15test_svtbl2_s6411svint64x2_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[DATA_COERCE0:%.*]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], <vscale x 2 x i64> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
//
svint64_t test_svtbl2_s64(svint64x2_t data, svuint64_t indices)
{
@ -93,17 +77,13 @@ svint64_t test_svtbl2_s64(svint64x2_t data, svuint64_t indices)
// CHECK-LABEL: @test_svtbl2_u8(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[DATA_COERCE0:%.*]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z14test_svtbl2_u811svuint8x2_tu11__SVUint8_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[DATA_COERCE0:%.*]], <vscale x 16 x i8> [[DATA_COERCE1:%.*]], <vscale x 16 x i8> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
//
svuint8_t test_svtbl2_u8(svuint8x2_t data, svuint8_t indices)
{
@ -112,17 +92,13 @@ svuint8_t test_svtbl2_u8(svuint8x2_t data, svuint8_t indices)
// CHECK-LABEL: @test_svtbl2_u16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[DATA_COERCE0:%.*]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], <vscale x 8 x i16> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z15test_svtbl2_u1612svuint16x2_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[DATA_COERCE0:%.*]], <vscale x 8 x i16> [[DATA_COERCE1:%.*]], <vscale x 8 x i16> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_svtbl2_u16(svuint16x2_t data, svuint16_t indices)
{
@ -131,17 +107,13 @@ svuint16_t test_svtbl2_u16(svuint16x2_t data, svuint16_t indices)
// CHECK-LABEL: @test_svtbl2_u32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[DATA_COERCE0:%.*]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], <vscale x 4 x i32> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z15test_svtbl2_u3212svuint32x2_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[DATA_COERCE0:%.*]], <vscale x 4 x i32> [[DATA_COERCE1:%.*]], <vscale x 4 x i32> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
//
svuint32_t test_svtbl2_u32(svuint32x2_t data, svuint32_t indices)
{
@ -150,17 +122,13 @@ svuint32_t test_svtbl2_u32(svuint32x2_t data, svuint32_t indices)
// CHECK-LABEL: @test_svtbl2_u64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[DATA_COERCE0:%.*]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], <vscale x 2 x i64> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z15test_svtbl2_u6412svuint64x2_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[DATA_COERCE0:%.*]], <vscale x 2 x i64> [[DATA_COERCE1:%.*]], <vscale x 2 x i64> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP0]]
//
svuint64_t test_svtbl2_u64(svuint64x2_t data, svuint64_t indices)
{
@ -169,17 +137,13 @@ svuint64_t test_svtbl2_u64(svuint64x2_t data, svuint64_t indices)
// CHECK-LABEL: @test_svtbl2_f16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tbl2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tbl2.nxv8f16(<vscale x 8 x half> [[DATA_COERCE0:%.*]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], <vscale x 8 x i16> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z15test_svtbl2_f1613svfloat16x2_tu12__SVUint16_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tbl2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tbl2.nxv8f16(<vscale x 8 x half> [[DATA_COERCE0:%.*]], <vscale x 8 x half> [[DATA_COERCE1:%.*]], <vscale x 8 x i16> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
//
svfloat16_t test_svtbl2_f16(svfloat16x2_t data, svuint16_t indices)
{
@ -188,17 +152,13 @@ svfloat16_t test_svtbl2_f16(svfloat16x2_t data, svuint16_t indices)
// CHECK-LABEL: @test_svtbl2_f32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tbl2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 4 x float> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tbl2.nxv4f32(<vscale x 4 x float> [[DATA_COERCE0:%.*]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], <vscale x 4 x i32> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z15test_svtbl2_f3213svfloat32x2_tu12__SVUint32_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tbl2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tbl2.nxv4f32(<vscale x 4 x float> [[DATA_COERCE0:%.*]], <vscale x 4 x float> [[DATA_COERCE1:%.*]], <vscale x 4 x i32> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
//
svfloat32_t test_svtbl2_f32(svfloat32x2_t data, svuint32_t indices)
{
@ -207,17 +167,13 @@ svfloat32_t test_svtbl2_f32(svfloat32x2_t data, svuint32_t indices)
// CHECK-LABEL: @test_svtbl2_f64(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tbl2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 2 x double> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tbl2.nxv2f64(<vscale x 2 x double> [[DATA_COERCE0:%.*]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], <vscale x 2 x i64> [[INDICES:%.*]])
// CHECK-NEXT: ret <vscale x 2 x double> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z15test_svtbl2_f6413svfloat64x2_tu12__SVUint64_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tbl2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 2 x double> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tbl2.nxv2f64(<vscale x 2 x double> [[DATA_COERCE0:%.*]], <vscale x 2 x double> [[DATA_COERCE1:%.*]], <vscale x 2 x i64> [[INDICES:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 2 x double> [[TMP0]]
//
svfloat64_t test_svtbl2_f64(svfloat64x2_t data, svuint64_t indices)
{

12
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_create2_bool.c
View File

@ -33,15 +33,15 @@
// CHECK-LABEL: @test_svcreate2_b(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> poison, <vscale x 16 x i1> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP0]], <vscale x 16 x i1> [[X1:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i1> [[TMP1]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[X1:%.*]], 1
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]]
//
// CPP-CHECK-LABEL: @_Z16test_svcreate2_bu10__SVBool_tS_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> poison, <vscale x 16 x i1> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP0]], <vscale x 16 x i1> [[X1:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i1> [[TMP1]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[X1:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]]
//
svboolx2_t test_svcreate2_b(svbool_t x0, svbool_t x1) MODE_ATTR
{

20
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_create4_bool.c
View File

@ -33,19 +33,19 @@
// CHECK-LABEL: @test_svcreate4_b(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> poison, <vscale x 16 x i1> [[X0:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TMP0]], <vscale x 16 x i1> [[X1:%.*]], i64 16)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TMP1]], <vscale x 16 x i1> [[X2:%.*]], i64 32)
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TMP2]], <vscale x 16 x i1> [[X4:%.*]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i1> [[TMP3]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[X0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[X1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[X2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[X4:%.*]], 3
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]]
//
// CPP-CHECK-LABEL: @_Z16test_svcreate4_bu10__SVBool_tS_S_S_(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> poison, <vscale x 16 x i1> [[X0:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TMP0]], <vscale x 16 x i1> [[X1:%.*]], i64 16)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TMP1]], <vscale x 16 x i1> [[X2:%.*]], i64 32)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TMP2]], <vscale x 16 x i1> [[X4:%.*]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i1> [[TMP3]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[X0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[X1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[X2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[X4:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]]
//
svboolx4_t test_svcreate4_b(svbool_t x0, svbool_t x1, svbool_t x2, svbool_t x4) ATTR
{

24
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_get2_bool.c
View File

@ -31,13 +31,17 @@
// CHECK-LABEL: @test_svget2_b_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.vector.extract.nxv16i1.nxv32i1(<vscale x 32 x i1> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], 0
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svget2_b_010svboolx2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.vector.extract.nxv16i1.nxv32i1(<vscale x 32 x i1> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], 0
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP2]]
//
svbool_t test_svget2_b_0(svboolx2_t tuple) ATTR
{
@ -46,13 +50,17 @@ svbool_t test_svget2_b_0(svboolx2_t tuple) ATTR
// CHECK-LABEL: @test_svget2_b_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.vector.extract.nxv16i1.nxv32i1(<vscale x 32 x i1> [[TUPLE:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], 1
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svget2_b_110svboolx2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.vector.extract.nxv16i1.nxv32i1(<vscale x 32 x i1> [[TUPLE:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], 1
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP2]]
//
svbool_t test_svget2_b_1(svboolx2_t tuple) ATTR
{

48
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_get4_bool.c
View File

@ -35,13 +35,21 @@
// several parameters, one for each member of the original struct.
// CHECK-LABEL: @test_svget4_b_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.vector.extract.nxv16i1.nxv64i1(<vscale x 64 x i1> [[TUPLE:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], 0
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_b_010svboolx4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.vector.extract.nxv16i1.nxv64i1(<vscale x 64 x i1> [[TUPLE:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], 0
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP4]]
//
svbool_t test_svget4_b_0(svboolx4_t tuple) MODE_ATTR
{
@ -52,13 +60,21 @@ svbool_t test_svget4_b_0(svboolx4_t tuple) MODE_ATTR
// several parameters, one for each member of the original struct.
// CHECK-LABEL: @test_svget4_b_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.vector.extract.nxv16i1.nxv64i1(<vscale x 64 x i1> [[TUPLE:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], 1
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_b_110svboolx4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.vector.extract.nxv16i1.nxv64i1(<vscale x 64 x i1> [[TUPLE:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], 1
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP4]]
//
svbool_t test_svget4_b_1(svboolx4_t tuple) MODE_ATTR
{
@ -69,13 +85,21 @@ svbool_t test_svget4_b_1(svboolx4_t tuple) MODE_ATTR
// several parameters, one for each member of the original struct.
// CHECK-LABEL: @test_svget4_b_3(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.vector.extract.nxv16i1.nxv64i1(<vscale x 64 x i1> [[TUPLE:%.*]], i64 48)
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], 3
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svget4_b_310svboolx4_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.vector.extract.nxv16i1.nxv64i1(<vscale x 64 x i1> [[TUPLE:%.*]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], 3
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP4]]
//
svbool_t test_svget4_b_3(svboolx4_t tuple) MODE_ATTR
{

704
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_ld1.c
View File

File diff suppressed because it is too large Load Diff

704
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_ldnt1.c
View File

File diff suppressed because it is too large Load Diff

1132
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_loads.c
View File

File diff suppressed because it is too large Load Diff

80
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_pext.c
View File

@ -141,21 +141,27 @@ svbool_t test_svpext_lane_c64_3(svcount_t c) ATTR {
// CHECK-LABEL: @test_svpext_lane_c8_x2_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i1>, <vscale x 16 x i1> } @llvm.aarch64.sve.pext.x2.nxv16i1(target("aarch64.svcount") [[C:%.*]], i32 0)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> poison, <vscale x 16 x i1> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP3]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i1> [[TMP4]]
// CHECK-NEXT: store <vscale x 32 x i1> [[TMP4]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svpext_lane_c8_x2_0u11__SVCount_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i1>, <vscale x 16 x i1> } @llvm.aarch64.sve.pext.x2.nxv16i1(target("aarch64.svcount") [[C:%.*]], i32 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> poison, <vscale x 16 x i1> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i1> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 32 x i1> [[TMP4]], ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP5]]
//
svboolx2_t test_svpext_lane_c8_x2_0(svcount_t c) ATTR {
return svpext_lane_c8_x2(c, 0);
@ -163,21 +169,27 @@ svboolx2_t test_svpext_lane_c8_x2_0(svcount_t c) ATTR {
// CHECK-LABEL: @test_svpext_lane_c8_x2_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i1>, <vscale x 16 x i1> } @llvm.aarch64.sve.pext.x2.nxv16i1(target("aarch64.svcount") [[C:%.*]], i32 1)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> poison, <vscale x 16 x i1> [[TMP1]], i64 0)
// CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], 1
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP3]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i1> [[TMP4]]
// CHECK-NEXT: store <vscale x 32 x i1> [[TMP4]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP5]]
//
// CPP-CHECK-LABEL: @_Z24test_svpext_lane_c8_x2_1u11__SVCount_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i1>, <vscale x 16 x i1> } @llvm.aarch64.sve.pext.x2.nxv16i1(target("aarch64.svcount") [[C:%.*]], i32 1)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> poison, <vscale x 16 x i1> [[TMP1]], i64 0)
// CPP-CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP3]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i1> [[TMP4]]
// CPP-CHECK-NEXT: store <vscale x 32 x i1> [[TMP4]], ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: [[TMP5:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP5]]
//
svboolx2_t test_svpext_lane_c8_x2_1(svcount_t c) ATTR {
return svpext_lane_c8_x2(c, 1);
@ -185,6 +197,7 @@ svboolx2_t test_svpext_lane_c8_x2_1(svcount_t c) ATTR {
// CHECK-LABEL: @test_svpext_lane_c16_x2_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i1>, <vscale x 8 x i1> } @llvm.aarch64.sve.pext.x2.nxv8i1(target("aarch64.svcount") [[C:%.*]], i32 0)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i1>, <vscale x 8 x i1> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP1]])
@ -192,10 +205,13 @@ svboolx2_t test_svpext_lane_c8_x2_1(svcount_t c) ATTR {
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i1>, <vscale x 8 x i1> } [[TMP0]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP4]])
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
// CPP-CHECK-LABEL: @_Z25test_svpext_lane_c16_x2_0u11__SVCount_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i1>, <vscale x 8 x i1> } @llvm.aarch64.sve.pext.x2.nxv8i1(target("aarch64.svcount") [[C:%.*]], i32 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i1>, <vscale x 8 x i1> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP1]])
@ -203,7 +219,9 @@ svboolx2_t test_svpext_lane_c8_x2_1(svcount_t c) ATTR {
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i1>, <vscale x 8 x i1> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP4]])
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CPP-CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
svboolx2_t test_svpext_lane_c16_x2_0(svcount_t c) ATTR {
return svpext_lane_c16_x2(c, 0);
@ -211,6 +229,7 @@ svboolx2_t test_svpext_lane_c16_x2_0(svcount_t c) ATTR {
// CHECK-LABEL: @test_svpext_lane_c16_x2_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i1>, <vscale x 8 x i1> } @llvm.aarch64.sve.pext.x2.nxv8i1(target("aarch64.svcount") [[C:%.*]], i32 1)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i1>, <vscale x 8 x i1> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP1]])
@ -218,10 +237,13 @@ svboolx2_t test_svpext_lane_c16_x2_0(svcount_t c) ATTR {
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i1>, <vscale x 8 x i1> } [[TMP0]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP4]])
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
// CPP-CHECK-LABEL: @_Z25test_svpext_lane_c16_x2_1u11__SVCount_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 8 x i1>, <vscale x 8 x i1> } @llvm.aarch64.sve.pext.x2.nxv8i1(target("aarch64.svcount") [[C:%.*]], i32 1)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 8 x i1>, <vscale x 8 x i1> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP1]])
@ -229,7 +251,9 @@ svboolx2_t test_svpext_lane_c16_x2_0(svcount_t c) ATTR {
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 8 x i1>, <vscale x 8 x i1> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP4]])
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CPP-CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
svboolx2_t test_svpext_lane_c16_x2_1(svcount_t c) ATTR {
return svpext_lane_c16_x2(c, 1);
@ -237,6 +261,7 @@ svboolx2_t test_svpext_lane_c16_x2_1(svcount_t c) ATTR {
// CHECK-LABEL: @test_svpext_lane_c32_x2_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i1>, <vscale x 4 x i1> } @llvm.aarch64.sve.pext.x2.nxv4i1(target("aarch64.svcount") [[C:%.*]], i32 0)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i1>, <vscale x 4 x i1> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP1]])
@ -244,10 +269,13 @@ svboolx2_t test_svpext_lane_c16_x2_1(svcount_t c) ATTR {
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i1>, <vscale x 4 x i1> } [[TMP0]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP4]])
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
// CPP-CHECK-LABEL: @_Z25test_svpext_lane_c32_x2_0u11__SVCount_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i1>, <vscale x 4 x i1> } @llvm.aarch64.sve.pext.x2.nxv4i1(target("aarch64.svcount") [[C:%.*]], i32 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i1>, <vscale x 4 x i1> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP1]])
@ -255,7 +283,9 @@ svboolx2_t test_svpext_lane_c16_x2_1(svcount_t c) ATTR {
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i1>, <vscale x 4 x i1> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP4]])
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CPP-CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
svboolx2_t test_svpext_lane_c32_x2_0(svcount_t c) ATTR {
return svpext_lane_c32_x2(c, 0);
@ -263,6 +293,7 @@ svboolx2_t test_svpext_lane_c32_x2_0(svcount_t c) ATTR {
// CHECK-LABEL: @test_svpext_lane_c32_x2_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i1>, <vscale x 4 x i1> } @llvm.aarch64.sve.pext.x2.nxv4i1(target("aarch64.svcount") [[C:%.*]], i32 1)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i1>, <vscale x 4 x i1> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP1]])
@ -270,10 +301,13 @@ svboolx2_t test_svpext_lane_c32_x2_0(svcount_t c) ATTR {
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i1>, <vscale x 4 x i1> } [[TMP0]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP4]])
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
// CPP-CHECK-LABEL: @_Z25test_svpext_lane_c32_x2_1u11__SVCount_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 4 x i1>, <vscale x 4 x i1> } @llvm.aarch64.sve.pext.x2.nxv4i1(target("aarch64.svcount") [[C:%.*]], i32 1)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 4 x i1>, <vscale x 4 x i1> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP1]])
@ -281,7 +315,9 @@ svboolx2_t test_svpext_lane_c32_x2_0(svcount_t c) ATTR {
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 4 x i1>, <vscale x 4 x i1> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP4]])
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CPP-CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
svboolx2_t test_svpext_lane_c32_x2_1(svcount_t c) ATTR {
return svpext_lane_c32_x2(c, 1);
@ -289,6 +325,7 @@ svboolx2_t test_svpext_lane_c32_x2_1(svcount_t c) ATTR {
// CHECK-LABEL: @test_svpext_lane_c64_x2_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i1>, <vscale x 2 x i1> } @llvm.aarch64.sve.pext.x2.nxv2i1(target("aarch64.svcount") [[C:%.*]], i32 0)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i1>, <vscale x 2 x i1> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP1]])
@ -296,10 +333,13 @@ svboolx2_t test_svpext_lane_c32_x2_1(svcount_t c) ATTR {
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i1>, <vscale x 2 x i1> } [[TMP0]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP4]])
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
// CPP-CHECK-LABEL: @_Z25test_svpext_lane_c64_x2_0u11__SVCount_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i1>, <vscale x 2 x i1> } @llvm.aarch64.sve.pext.x2.nxv2i1(target("aarch64.svcount") [[C:%.*]], i32 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i1>, <vscale x 2 x i1> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP1]])
@ -307,7 +347,9 @@ svboolx2_t test_svpext_lane_c32_x2_1(svcount_t c) ATTR {
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i1>, <vscale x 2 x i1> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP4]])
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CPP-CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
svboolx2_t test_svpext_lane_c64_x2_0(svcount_t c) ATTR {
return svpext_lane_c64_x2(c, 0);
@ -315,6 +357,7 @@ svboolx2_t test_svpext_lane_c64_x2_0(svcount_t c) ATTR {
// CHECK-LABEL: @test_svpext_lane_c64_x2_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i1>, <vscale x 2 x i1> } @llvm.aarch64.sve.pext.x2.nxv2i1(target("aarch64.svcount") [[C:%.*]], i32 1)
// CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i1>, <vscale x 2 x i1> } [[TMP0]], 0
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP1]])
@ -322,10 +365,13 @@ svboolx2_t test_svpext_lane_c64_x2_0(svcount_t c) ATTR {
// CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i1>, <vscale x 2 x i1> } [[TMP0]], 1
// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP4]])
// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
// CPP-CHECK-LABEL: @_Z25test_svpext_lane_c64_x2_1u11__SVCount_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[RETVAL:%.*]] = alloca { <vscale x 16 x i1>, <vscale x 16 x i1> }, align 8
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 2 x i1>, <vscale x 2 x i1> } @llvm.aarch64.sve.pext.x2.nxv2i1(target("aarch64.svcount") [[C:%.*]], i32 1)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = extractvalue { <vscale x 2 x i1>, <vscale x 2 x i1> } [[TMP0]], 0
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP1]])
@ -333,7 +379,9 @@ svboolx2_t test_svpext_lane_c64_x2_0(svcount_t c) ATTR {
// CPP-CHECK-NEXT: [[TMP4:%.*]] = extractvalue { <vscale x 2 x i1>, <vscale x 2 x i1> } [[TMP0]], 1
// CPP-CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP4]])
// CPP-CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TMP3]], <vscale x 16 x i1> [[TMP5]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i1> [[TMP6]]
// CPP-CHECK-NEXT: store <vscale x 32 x i1> [[TMP6]], ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: [[TMP7:%.*]] = load { <vscale x 16 x i1>, <vscale x 16 x i1> }, ptr [[RETVAL]], align 8
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP7]]
//
svboolx2_t test_svpext_lane_c64_x2_1(svcount_t c) ATTR {
return svpext_lane_c64_x2(c, 1);

36
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_qcvtn.c
View File

@ -29,17 +29,13 @@
// CHECK-LABEL: @test_qcvtn_s16_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtn.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtn.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z21test_qcvtn_s16_s32_x211svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtn.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtn.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svint16_t test_qcvtn_s16_s32_x2(svint32x2_t zn) ATTR {
return SVE_ACLE_FUNC(svqcvtn_s16,_s32_x2,,)(zn);
@ -47,17 +43,13 @@ svint16_t test_qcvtn_s16_s32_x2(svint32x2_t zn) ATTR {
// CHECK-LABEL: @test_qcvtn_u16_u32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvtn.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvtn.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z21test_qcvtn_u16_u32_x212svuint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvtn.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvtn.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_qcvtn_u16_u32_x2(svuint32x2_t zn) ATTR {
return SVE_ACLE_FUNC(svqcvtn_u16,_u32_x2,,)(zn);
@ -65,17 +57,13 @@ svuint16_t test_qcvtn_u16_u32_x2(svuint32x2_t zn) ATTR {
// CHECK-LABEL: @test_qcvtn_u16_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtun.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtun.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z21test_qcvtn_u16_s32_x211svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtun.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtun.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]])
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_qcvtn_u16_s32_x2(svint32x2_t zn) ATTR {
return SVE_ACLE_FUNC(svqcvtn_u16,_s32_x2,,)(zn);

36
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_qrshr.c
View File

@ -28,17 +28,13 @@
// CHECK-LABEL: @test_svqrshrn_s16_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrn.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrn.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svqrshrn_s16_s32_x211svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrn.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrn.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svint16_t test_svqrshrn_s16_s32_x2(svint32x2_t zn) ATTR {
return SVE_ACLE_FUNC(svqrshrn,_n,_s16,_s32_x2,)(zn, 16);
@ -48,17 +44,13 @@ svint16_t test_svqrshrn_s16_s32_x2(svint32x2_t zn) ATTR {
// CHECK-LABEL: @test_svqrshrn_u16_u32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshrn.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshrn.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z24test_svqrshrn_u16_u32_x212svuint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshrn.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.uqrshrn.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_svqrshrn_u16_u32_x2(svuint32x2_t zn) ATTR {
return SVE_ACLE_FUNC(svqrshrn,_n,_u16,_u32_x2,)(zn, 16);
@ -68,17 +60,13 @@ svuint16_t test_svqrshrn_u16_u32_x2(svuint32x2_t zn) ATTR {
// CHECK-LABEL: @test_svqrshrun_u16_s32_x2(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrun.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrun.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
// CPP-CHECK-LABEL: @_Z25test_svqrshrun_u16_s32_x211svint32x2_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN:%.*]], i64 0)
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[ZN]], i64 4)
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrun.x2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP2]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.sqrshrun.x2.nxv4i32(<vscale x 4 x i32> [[ZN_COERCE0:%.*]], <vscale x 4 x i32> [[ZN_COERCE1:%.*]], i32 16)
// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
//
svuint16_t test_svqrshrun_u16_s32_x2(svint32x2_t zn) ATTR {
return SVE_ACLE_FUNC(svqrshrun,_n,_u16,_s32_x2,)(zn, 16);

24
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_set2_bool.c
View File

@ -33,13 +33,17 @@
// CHECK-LABEL: @test_svset2_b_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TUPLE:%.*]], <vscale x 16 x i1> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 32 x i1> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svset2_b_010svboolx2_tu10__SVBool_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TUPLE:%.*]], <vscale x 16 x i1> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 32 x i1> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]]
//
svboolx2_t test_svset2_b_0(svboolx2_t tuple, svbool_t x) ATTR
{
@ -48,13 +52,17 @@ svboolx2_t test_svset2_b_0(svboolx2_t tuple, svbool_t x) ATTR
// CHECK-LABEL: @test_svset2_b_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TUPLE:%.*]], <vscale x 16 x i1> [[X:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 32 x i1> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]]
//
// CPP-CHECK-LABEL: @_Z15test_svset2_b_110svboolx2_tu10__SVBool_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 32 x i1> @llvm.vector.insert.nxv32i1.nxv16i1(<vscale x 32 x i1> [[TUPLE:%.*]], <vscale x 16 x i1> [[X:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 32 x i1> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]]
//
svboolx2_t test_svset2_b_1(svboolx2_t tuple, svbool_t x) ATTR
{

48
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_set4_bool.c
View File

@ -32,13 +32,21 @@
// CHECK-LABEL: @test_svset4_b_0(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TUPLE:%.*]], <vscale x 16 x i1> [[X:%.*]], i64 0)
// CHECK-NEXT: ret <vscale x 64 x i1> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], <vscale x 16 x i1> [[X:%.*]], 0
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_b_010svboolx4_tu10__SVBool_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TUPLE:%.*]], <vscale x 16 x i1> [[X:%.*]], i64 0)
// CPP-CHECK-NEXT: ret <vscale x 64 x i1> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], <vscale x 16 x i1> [[X:%.*]], 0
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP4]]
//
svboolx4_t test_svset4_b_0(svboolx4_t tuple, svbool_t x) ATTR
{
@ -47,13 +55,21 @@ svboolx4_t test_svset4_b_0(svboolx4_t tuple, svbool_t x) ATTR
// CHECK-LABEL: @test_svset4_b_1(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TUPLE:%.*]], <vscale x 16 x i1> [[X:%.*]], i64 16)
// CHECK-NEXT: ret <vscale x 64 x i1> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], <vscale x 16 x i1> [[X:%.*]], 1
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_b_110svboolx4_tu10__SVBool_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TUPLE:%.*]], <vscale x 16 x i1> [[X:%.*]], i64 16)
// CPP-CHECK-NEXT: ret <vscale x 64 x i1> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], <vscale x 16 x i1> [[X:%.*]], 1
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP4]]
//
svboolx4_t test_svset4_b_1(svboolx4_t tuple, svbool_t x) ATTR
{
@ -62,13 +78,21 @@ svboolx4_t test_svset4_b_1(svboolx4_t tuple, svbool_t x) ATTR
// CHECK-LABEL: @test_svset4_b_3(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TUPLE:%.*]], <vscale x 16 x i1> [[X:%.*]], i64 48)
// CHECK-NEXT: ret <vscale x 64 x i1> [[TMP0]]
// CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], <vscale x 16 x i1> [[X:%.*]], 3
// CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP4]]
//
// CPP-CHECK-LABEL: @_Z15test_svset4_b_310svboolx4_tu10__SVBool_t(
// CPP-CHECK-NEXT: entry:
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 64 x i1> @llvm.vector.insert.nxv64i1.nxv16i1(<vscale x 64 x i1> [[TUPLE:%.*]], <vscale x 16 x i1> [[X:%.*]], i64 48)
// CPP-CHECK-NEXT: ret <vscale x 64 x i1> [[TMP0]]
// CPP-CHECK-NEXT: [[TMP0:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } poison, <vscale x 16 x i1> [[TUPLE_COERCE0:%.*]], 0
// CPP-CHECK-NEXT: [[TMP1:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP0]], <vscale x 16 x i1> [[TUPLE_COERCE1:%.*]], 1
// CPP-CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP1]], <vscale x 16 x i1> [[TUPLE_COERCE2:%.*]], 2
// CPP-CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP2]], <vscale x 16 x i1> [[TUPLE_COERCE3:%.*]], 3
// CPP-CHECK-NEXT: [[TMP4:%.*]] = insertvalue { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP3]], <vscale x 16 x i1> [[X:%.*]], 3
// CPP-CHECK-NEXT: ret { <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1> } [[TMP4]]
//
svboolx4_t test_svset4_b_3(svboolx4_t tuple, svbool_t x) ATTR
{

792
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_st1.c
View File

File diff suppressed because it is too large Load Diff

792
clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_stnt1.c
View File

File diff suppressed because it is too large Load Diff

Some files were not shown because too many files have changed in this diff Show More