[SelectionDAG] Deal with POISON for INSERT_VECTOR_ELT/INSERT_SUBVECTOR

As reported in #141034 SelectionDAG::getNode had some unexpected
behaviors when trying to create vectors with UNDEF elements. Since
we treat both UNDEF and POISON as undefined (when using isUndef())
we can't just fold away INSERT_VECTOR_ELT/INSERT_SUBVECTOR based on
isUndef(), as that could make the resulting vector more poisonous.

Same kind of bug existed in DAGCombiner::visitINSERT_SUBVECTOR.

Here are some examples:

This fold was done even if vec[idx] was POISON:
  INSERT_VECTOR_ELT vec, UNDEF, idx -> vec

This fold was done even if any of vec[idx..idx+size] was POISON:
  INSERT_SUBVECTOR vec, UNDEF, idx -> vec

This fold was done even if the elements not extracted from vec could
be POISON:
  sub = EXTRACT_SUBVECTOR vec, idx
  INSERT_SUBVECTOR UNDEF, sub, idx -> vec

With this patch we avoid such folds unless we can prove that the
result isn't more poisonous when eliminating the insert.

Fixes #141034

llvm/include/llvm/CodeGen/SelectionDAGNodes.h

@@ -1885,6 +1885,12 @@ LLVM_ABI SDValue peekThroughOneUseBitcasts(SDValue V);
 /// If \p V is not an extracted subvector, it is returned as-is.
 LLVM_ABI SDValue peekThroughExtractSubvectors(SDValue V);
 
+/// Recursively peek through INSERT_VECTOR_ELT nodes, returning the source
+/// vector operand of \p V, as long as \p V is an INSERT_VECTOR_ELT operation
+/// that do not insert into any of the demanded vector elts.
+LLVM_ABI SDValue peekThroughInsertVectorElt(SDValue V,
+                                            const APInt &DemandedElts);
+
 /// Return the non-truncated source operand of \p V if it exists.
 /// If \p V is not a truncation, it is returned as-is.
 LLVM_ABI SDValue peekThroughTruncates(SDValue V);

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

@@ -23298,6 +23298,7 @@ SDValue DAGCombiner::visitINSERT_VECTOR_ELT(SDNode *N) {
   auto *IndexC = dyn_cast<ConstantSDNode>(EltNo);
 
   // Insert into out-of-bounds element is undefined.
+  // Code below relies on that we handle this special case early.
   if (IndexC && VT.isFixedLengthVector() &&
       IndexC->getZExtValue() >= VT.getVectorNumElements())
     return DAG.getUNDEF(VT);
@@ -23308,14 +23309,28 @@ SDValue DAGCombiner::visitINSERT_VECTOR_ELT(SDNode *N) {
       InVec == InVal.getOperand(0) && EltNo == InVal.getOperand(1))
     return InVec;
 
-  if (!IndexC) {
-    // If this is variable insert to undef vector, it might be better to splat:
-    // inselt undef, InVal, EltNo --> build_vector < InVal, InVal, ... >
-    if (InVec.isUndef() && TLI.shouldSplatInsEltVarIndex(VT))
-      return DAG.getSplat(VT, DL, InVal);
-    return SDValue();
+  // If this is variable insert to undef vector, it might be better to splat:
+  // inselt undef, InVal, EltNo --> build_vector < InVal, InVal, ... >
+  if (!IndexC && InVec.isUndef() && TLI.shouldSplatInsEltVarIndex(VT))
+    return DAG.getSplat(VT, DL, InVal);
+
+  // Try to drop insert of UNDEF/POISON elements. This is also done in getNode,
+  // but we also do it as a DAG combine since for example simplifications into
+  // SPLAT_VECTOR/BUILD_VECTOR may turn poison elements into undef/zero etc, and
+  // then suddenly the InVec is guaranteed to not be poison.
+  if (InVal.isUndef()) {
+    if (IndexC && VT.isFixedLengthVector()) {
+      APInt EltMask = APInt::getOneBitSet(VT.getVectorNumElements(),
+                                          IndexC->getZExtValue());
+      if (DAG.isGuaranteedNotToBePoison(InVec, EltMask))
+        return InVec;
+    }
+    return DAG.getFreeze(InVec);
   }
 
+  if (!IndexC)
+    return SDValue();
+
   if (VT.isScalableVector())
     return SDValue();
 
@@ -27799,18 +27814,42 @@ SDValue DAGCombiner::visitINSERT_SUBVECTOR(SDNode *N) {
   SDValue N2 = N->getOperand(2);
   uint64_t InsIdx = N->getConstantOperandVal(2);
 
-  // If inserting an UNDEF, just return the original vector.
-  if (N1.isUndef())
-    return N0;
+  // If inserting an UNDEF, just return the original vector (unless it makes the
+  // result more poisonous).
+  if (N1.isUndef()) {
+    if (N1.getOpcode() == ISD::POISON)
+      return N0;
+    if (VT.isFixedLengthVector()) {
+      unsigned SubVecNumElts = N1.getValueType().getVectorNumElements();
+      APInt EltMask = APInt::getBitsSet(VT.getVectorNumElements(), InsIdx,
+                                        InsIdx + SubVecNumElts);
+      if (DAG.isGuaranteedNotToBePoison(N0, EltMask))
+        return N0;
+    }
+    return DAG.getFreeze(N0);
+  }
 
-  // If this is an insert of an extracted vector into an undef vector, we can
-  // just use the input to the extract if the types match, and can simplify
+  // If this is an insert of an extracted vector into an undef/poison vector, we
+  // can just use the input to the extract if the types match, and can simplify
   // in some cases even if they don't.
   if (N0.isUndef() && N1.getOpcode() == ISD::EXTRACT_SUBVECTOR &&
       N1.getOperand(1) == N2) {
+    EVT N1VT = N1.getValueType();
     EVT SrcVT = N1.getOperand(0).getValueType();
-    if (SrcVT == VT)
-      return N1.getOperand(0);
+    if (SrcVT == VT) {
+      // Need to ensure that result isn't more poisonous if skipping both the
+      // extract+insert.
+      if (N0.getOpcode() == ISD::POISON)
+        return N1.getOperand(0);
+      if (VT.isFixedLengthVector() && N1VT.isFixedLengthVector()) {
+        unsigned SubVecNumElts = N1VT.getVectorNumElements();
+        APInt EltMask = APInt::getBitsSet(VT.getVectorNumElements(), InsIdx,
+                                          InsIdx + SubVecNumElts);
+        if (DAG.isGuaranteedNotToBePoison(N1.getOperand(0), ~EltMask))
+          return N1.getOperand(0);
+      } else if (DAG.isGuaranteedNotToBePoison(N1.getOperand(0)))
+        return N1.getOperand(0);
+    }
     // TODO: To remove the zero check, need to adjust the offset to
     // a multiple of the new src type.
     if (isNullConstant(N2)) {

llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

@@ -5513,8 +5513,9 @@ bool SelectionDAG::isGuaranteedNotToBeUndefOrPoison(SDValue Op,
       APInt InVecDemandedElts = DemandedElts;
       InVecDemandedElts.clearBit(IndexC->getZExtValue());
       if (!!InVecDemandedElts &&
-          !isGuaranteedNotToBeUndefOrPoison(InVec, InVecDemandedElts,
-                                            PoisonOnly, Depth + 1))
+          !isGuaranteedNotToBeUndefOrPoison(
+              peekThroughInsertVectorElt(InVec, InVecDemandedElts),
+              InVecDemandedElts, PoisonOnly, Depth + 1))
         return false;
       return true;
     }
@@ -8215,23 +8216,42 @@ SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
     // INSERT_VECTOR_ELT into out-of-bounds element is an UNDEF, except
     // for scalable vectors where we will generate appropriate code to
     // deal with out-of-bounds cases correctly.
-    if (N3C && N1.getValueType().isFixedLengthVector() &&
-        N3C->getZExtValue() >= N1.getValueType().getVectorNumElements())
+    if (N3C && VT.isFixedLengthVector() &&
+        N3C->getZExtValue() >= VT.getVectorNumElements())
       return getUNDEF(VT);
 
     // Undefined index can be assumed out-of-bounds, so that's UNDEF too.
     if (N3.isUndef())
       return getUNDEF(VT);
 
-    // If the inserted element is an UNDEF, just use the input vector.
-    if (N2.isUndef())
+    // If inserting poison, just use the input vector.
+    if (N2.getOpcode() == ISD::POISON)
       return N1;
 
+    // Inserting undef into undef/poison is still undef.
+    if (N2.getOpcode() == ISD::UNDEF && N1.isUndef())
+      return getUNDEF(VT);
+
+    // If the inserted element is an UNDEF, just use the input vector.
+    // But not if skipping the insert could make the result more poisonous.
+    if (N2.isUndef()) {
+      if (N3C && VT.isFixedLengthVector()) {
+        APInt EltMask =
+            APInt::getOneBitSet(VT.getVectorNumElements(), N3C->getZExtValue());
+        if (isGuaranteedNotToBePoison(N1, EltMask))
+          return N1;
+      } else if (isGuaranteedNotToBePoison(N1))
+        return N1;
+    }
     break;
   }
   case ISD::INSERT_SUBVECTOR: {
-    // Inserting undef into undef is still undef.
-    if (N1.isUndef() && N2.isUndef())
+    // If inserting poison, just use the input vector,
+    if (N2.getOpcode() == ISD::POISON)
+      return N1;
+
+    // Inserting undef into undef/poison is still undef.
+    if (N2.getOpcode() == ISD::UNDEF && N1.isUndef())
       return getUNDEF(VT);
 
     EVT N2VT = N2.getValueType();
@@ -8260,11 +8280,37 @@ SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
     if (VT == N2VT)
       return N2;
 
-    // If this is an insert of an extracted vector into an undef vector, we
-    // can just use the input to the extract.
+    // If this is an insert of an extracted vector into an undef/poison vector,
+    // we can just use the input to the extract. But not if skipping the
+    // extract+insert could make the result more poisonous.
     if (N1.isUndef() && N2.getOpcode() == ISD::EXTRACT_SUBVECTOR &&
-        N2.getOperand(1) == N3 && N2.getOperand(0).getValueType() == VT)
-      return N2.getOperand(0);
+        N2.getOperand(1) == N3 && N2.getOperand(0).getValueType() == VT) {
+      if (N1.getOpcode() == ISD::POISON)
+        return N2.getOperand(0);
+      if (VT.isFixedLengthVector() && N2VT.isFixedLengthVector()) {
+        unsigned LoBit = N3->getAsZExtVal();
+        unsigned HiBit = LoBit + N2VT.getVectorNumElements();
+        APInt EltMask =
+            APInt::getBitsSet(VT.getVectorNumElements(), LoBit, HiBit);
+        if (isGuaranteedNotToBePoison(N2.getOperand(0), ~EltMask))
+          return N2.getOperand(0);
+      } else if (isGuaranteedNotToBePoison(N2.getOperand(0)))
+        return N2.getOperand(0);
+    }
+
+    // If the inserted subvector is UNDEF, just use the input vector.
+    // But not if skipping the insert could make the result more poisonous.
+    if (N2.isUndef()) {
+      if (VT.isFixedLengthVector()) {
+        unsigned LoBit = N3->getAsZExtVal();
+        unsigned HiBit = LoBit + N2VT.getVectorNumElements();
+        APInt EltMask =
+            APInt::getBitsSet(VT.getVectorNumElements(), LoBit, HiBit);
+        if (isGuaranteedNotToBePoison(N1, EltMask))
+          return N1;
+      } else if (isGuaranteedNotToBePoison(N1))
+        return N1;
+    }
     break;
   }
   case ISD::BITCAST:
@@ -12729,6 +12775,23 @@ SDValue llvm::peekThroughExtractSubvectors(SDValue V) {
   return V;
 }
 
+SDValue llvm::peekThroughInsertVectorElt(SDValue V, const APInt &DemandedElts) {
+  while (V.getOpcode() == ISD::INSERT_VECTOR_ELT) {
+    SDValue InVec = V.getOperand(0);
+    SDValue EltNo = V.getOperand(2);
+    EVT VT = InVec.getValueType();
+    auto *IndexC = dyn_cast<ConstantSDNode>(EltNo);
+    if (IndexC && VT.isFixedLengthVector() &&
+        IndexC->getAPIntValue().ult(VT.getVectorNumElements()) &&
+        !DemandedElts[IndexC->getZExtValue()]) {
+      V = InVec;
+      continue;
+    }
+    break;
+  }
+  return V;
+}
+
 SDValue llvm::peekThroughTruncates(SDValue V) {
   while (V.getOpcode() == ISD::TRUNCATE)
     V = V.getOperand(0);

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

@@ -3446,8 +3446,8 @@ bool TargetLowering::SimplifyDemandedVectorElts(
     break;
   }
   case ISD::INSERT_SUBVECTOR: {
-    // Demand any elements from the subvector and the remainder from the src its
-    // inserted into.
+    // Demand any elements from the subvector and the remainder from the src it
+    // is inserted into.
     SDValue Src = Op.getOperand(0);
     SDValue Sub = Op.getOperand(1);
     uint64_t Idx = Op.getConstantOperandVal(2);
@@ -3456,6 +3456,10 @@ bool TargetLowering::SimplifyDemandedVectorElts(
     APInt DemandedSrcElts = DemandedElts;
     DemandedSrcElts.clearBits(Idx, Idx + NumSubElts);
 
+    // If none of the sub operand elements are demanded, bypass the insert.
+    if (!DemandedSubElts)
+      return TLO.CombineTo(Op, Src);
+
     APInt SubUndef, SubZero;
     if (SimplifyDemandedVectorElts(Sub, DemandedSubElts, SubUndef, SubZero, TLO,
                                    Depth + 1))

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

@@ -15357,7 +15357,7 @@ SDValue AArch64TargetLowering::LowerBUILD_VECTOR(SDValue Op,
     for (unsigned i = 0; i < NumElts; ++i) {
       SDValue V = Op.getOperand(i);
       SDValue LaneIdx = DAG.getConstant(i, DL, MVT::i64);
-      if (!isIntOrFPConstant(V))
+      if (!isIntOrFPConstant(V) && !V.isUndef())
         // Note that type legalization likely mucked about with the VT of the
         // source operand, so we may have to convert it here before inserting.
         Val = DAG.getNode(ISD::INSERT_VECTOR_ELT, DL, VT, Val, V, LaneIdx);

llvm/test/CodeGen/AArch64/concat-vector-add-combine.ll

@@ -94,16 +94,14 @@ define i32 @combine_undef_add_8xi32(i32 %a, i32 %b, i32 %c, i32 %d) local_unname
 ; CHECK-LABEL: combine_undef_add_8xi32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    fmov s1, w0
-; CHECK-NEXT:    movi v0.2d, #0000000000000000
+; CHECK-NEXT:    dup v0.4s, w8
 ; CHECK-NEXT:    mov v1.s[1], w1
-; CHECK-NEXT:    uhadd v0.4h, v0.4h, v0.4h
 ; CHECK-NEXT:    mov v1.s[2], w2
 ; CHECK-NEXT:    mov v1.s[3], w3
-; CHECK-NEXT:    xtn v2.4h, v1.4s
-; CHECK-NEXT:    shrn v1.4h, v1.4s, #16
-; CHECK-NEXT:    uhadd v1.4h, v2.4h, v1.4h
-; CHECK-NEXT:    mov v1.d[1], v0.d[0]
-; CHECK-NEXT:    uaddlv s0, v1.8h
+; CHECK-NEXT:    uzp2 v2.8h, v1.8h, v0.8h
+; CHECK-NEXT:    uzp1 v0.8h, v1.8h, v0.8h
+; CHECK-NEXT:    uhadd v0.8h, v0.8h, v2.8h
+; CHECK-NEXT:    uaddlv s0, v0.8h
 ; CHECK-NEXT:    fmov w0, s0
 ; CHECK-NEXT:    ret
   %a1 = insertelement <8 x i32> poison, i32 %a, i32 0

llvm/test/CodeGen/AArch64/sve-fixed-length-masked-gather.ll

@@ -1198,11 +1198,15 @@ define void @masked_gather_passthru(ptr %a, ptr %b, ptr %c) vscale_range(16,0) #
 ; CHECK-NEXT:    ptrue p0.s, vl32
 ; CHECK-NEXT:    ptrue p2.d, vl32
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
-; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x2]
 ; CHECK-NEXT:    fcmeq p1.s, p0/z, z0.s, #0.0
 ; CHECK-NEXT:    ld1d { z0.d }, p2/z, [x1]
 ; CHECK-NEXT:    punpklo p2.h, p1.b
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    ld1w { z0.d }, p2/z, [z0.d]
+; CHECK-NEXT:    and z1.s, z1.s, #0x1
+; CHECK-NEXT:    cmpne p1.s, p1/z, z1.s, #0
+; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x2]
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    sel z0.s, p1, z0.s, z1.s
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vfw-web-simplification.ll

@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=riscv64 -mattr=+v,+zvfh,+f,+d -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=1 | FileCheck %s --check-prefixes=NO_FOLDING
-; RUN: llc -mtriple=riscv64 -mattr=+v,+zvfh,+f,+d -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=2 | FileCheck %s --check-prefixes=NO_FOLDING
+; RUN: llc -mtriple=riscv64 -mattr=+v,+zvfh,+f,+d -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=1 | FileCheck %s --check-prefixes=NO_FOLDING,NO_FOLDING1
+; RUN: llc -mtriple=riscv64 -mattr=+v,+zvfh,+f,+d -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=2 | FileCheck %s --check-prefixes=NO_FOLDING,NO_FOLDING2
 ; RUN: llc -mtriple=riscv64 -mattr=+v,+zvfh,+f,+d -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=3 | FileCheck %s --check-prefixes=FOLDING,ZVFH
 ; RUN: llc -mtriple=riscv64 -mattr=+v,+zvfhmin,+f,+d -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=3 | FileCheck %s --check-prefixes=FOLDING,ZVFHMIN
 ; Check that the default value enables the web folding and
@@ -8,20 +8,35 @@
 ; RUN: llc -mtriple=riscv64 -mattr=+v,+zvfh,+f,+d -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=FOLDING
 
 define void @vfwmul_v2f116_multiple_users(ptr %x, ptr %y, ptr %z, <2 x half> %a, <2 x half> %b, <2 x half> %b2) {
-; NO_FOLDING-LABEL: vfwmul_v2f116_multiple_users:
-; NO_FOLDING:       # %bb.0:
-; NO_FOLDING-NEXT:    vsetivli zero, 2, e16, mf4, ta, ma
-; NO_FOLDING-NEXT:    vfwcvt.f.f.v v11, v8
-; NO_FOLDING-NEXT:    vfwcvt.f.f.v v8, v9
-; NO_FOLDING-NEXT:    vfwcvt.f.f.v v9, v10
-; NO_FOLDING-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
-; NO_FOLDING-NEXT:    vfmul.vv v10, v11, v8
-; NO_FOLDING-NEXT:    vfadd.vv v11, v11, v9
-; NO_FOLDING-NEXT:    vfsub.vv v8, v8, v9
-; NO_FOLDING-NEXT:    vse32.v v10, (a0)
-; NO_FOLDING-NEXT:    vse32.v v11, (a1)
-; NO_FOLDING-NEXT:    vse32.v v8, (a2)
-; NO_FOLDING-NEXT:    ret
+; NO_FOLDING1-LABEL: vfwmul_v2f116_multiple_users:
+; NO_FOLDING1:       # %bb.0:
+; NO_FOLDING1-NEXT:    vsetivli zero, 2, e16, mf4, ta, ma
+; NO_FOLDING1-NEXT:    vfwcvt.f.f.v v11, v8
+; NO_FOLDING1-NEXT:    vfwcvt.f.f.v v8, v9
+; NO_FOLDING1-NEXT:    vfwcvt.f.f.v v9, v10
+; NO_FOLDING1-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; NO_FOLDING1-NEXT:    vfmul.vv v10, v11, v8
+; NO_FOLDING1-NEXT:    vfadd.vv v11, v11, v9
+; NO_FOLDING1-NEXT:    vfsub.vv v8, v8, v9
+; NO_FOLDING1-NEXT:    vse32.v v10, (a0)
+; NO_FOLDING1-NEXT:    vse32.v v11, (a1)
+; NO_FOLDING1-NEXT:    vse32.v v8, (a2)
+; NO_FOLDING1-NEXT:    ret
+;
+; NO_FOLDING2-LABEL: vfwmul_v2f116_multiple_users:
+; NO_FOLDING2:       # %bb.0:
+; NO_FOLDING2-NEXT:    vsetivli zero, 2, e16, mf4, ta, ma
+; NO_FOLDING2-NEXT:    vfwcvt.f.f.v v11, v8
+; NO_FOLDING2-NEXT:    vfwcvt.f.f.v v8, v9
+; NO_FOLDING2-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; NO_FOLDING2-NEXT:    vfmul.vv v9, v11, v8
+; NO_FOLDING2-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; NO_FOLDING2-NEXT:    vfwadd.wv v11, v11, v10
+; NO_FOLDING2-NEXT:    vfwsub.wv v8, v8, v10
+; NO_FOLDING2-NEXT:    vse32.v v9, (a0)
+; NO_FOLDING2-NEXT:    vse32.v v11, (a1)
+; NO_FOLDING2-NEXT:    vse32.v v8, (a2)
+; NO_FOLDING2-NEXT:    ret
 ;
 ; ZVFH-LABEL: vfwmul_v2f116_multiple_users:
 ; ZVFH:       # %bb.0:
@@ -61,20 +76,35 @@ define void @vfwmul_v2f116_multiple_users(ptr %x, ptr %y, ptr %z, <2 x half> %a,
 }
 
 define void @vfwmul_v2f32_multiple_users(ptr %x, ptr %y, ptr %z, <2 x float> %a, <2 x float> %b, <2 x float> %b2) {
-; NO_FOLDING-LABEL: vfwmul_v2f32_multiple_users:
-; NO_FOLDING:       # %bb.0:
-; NO_FOLDING-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
-; NO_FOLDING-NEXT:    vfwcvt.f.f.v v11, v8
-; NO_FOLDING-NEXT:    vfwcvt.f.f.v v8, v9
-; NO_FOLDING-NEXT:    vfwcvt.f.f.v v9, v10
-; NO_FOLDING-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
-; NO_FOLDING-NEXT:    vfmul.vv v10, v11, v8
-; NO_FOLDING-NEXT:    vfadd.vv v11, v11, v9
-; NO_FOLDING-NEXT:    vfsub.vv v8, v8, v9
-; NO_FOLDING-NEXT:    vse64.v v10, (a0)
-; NO_FOLDING-NEXT:    vse64.v v11, (a1)
-; NO_FOLDING-NEXT:    vse64.v v8, (a2)
-; NO_FOLDING-NEXT:    ret
+; NO_FOLDING1-LABEL: vfwmul_v2f32_multiple_users:
+; NO_FOLDING1:       # %bb.0:
+; NO_FOLDING1-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
+; NO_FOLDING1-NEXT:    vfwcvt.f.f.v v11, v8
+; NO_FOLDING1-NEXT:    vfwcvt.f.f.v v8, v9
+; NO_FOLDING1-NEXT:    vfwcvt.f.f.v v9, v10
+; NO_FOLDING1-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; NO_FOLDING1-NEXT:    vfmul.vv v10, v11, v8
+; NO_FOLDING1-NEXT:    vfadd.vv v11, v11, v9
+; NO_FOLDING1-NEXT:    vfsub.vv v8, v8, v9
+; NO_FOLDING1-NEXT:    vse64.v v10, (a0)
+; NO_FOLDING1-NEXT:    vse64.v v11, (a1)
+; NO_FOLDING1-NEXT:    vse64.v v8, (a2)
+; NO_FOLDING1-NEXT:    ret
+;
+; NO_FOLDING2-LABEL: vfwmul_v2f32_multiple_users:
+; NO_FOLDING2:       # %bb.0:
+; NO_FOLDING2-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
+; NO_FOLDING2-NEXT:    vfwcvt.f.f.v v11, v8
+; NO_FOLDING2-NEXT:    vfwcvt.f.f.v v8, v9
+; NO_FOLDING2-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; NO_FOLDING2-NEXT:    vfmul.vv v9, v11, v8
+; NO_FOLDING2-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; NO_FOLDING2-NEXT:    vfwadd.wv v11, v11, v10
+; NO_FOLDING2-NEXT:    vfwsub.wv v8, v8, v10
+; NO_FOLDING2-NEXT:    vse64.v v9, (a0)
+; NO_FOLDING2-NEXT:    vse64.v v11, (a1)
+; NO_FOLDING2-NEXT:    vse64.v v8, (a2)
+; NO_FOLDING2-NEXT:    ret
 ;
 ; FOLDING-LABEL: vfwmul_v2f32_multiple_users:
 ; FOLDING:       # %bb.0:

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-vw-web-simplification.ll

@@ -1,8 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=riscv32 -mattr=+v -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=1 | FileCheck %s --check-prefixes=NO_FOLDING
-; RUN: llc -mtriple=riscv64 -mattr=+v -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=1 | FileCheck %s --check-prefixes=NO_FOLDING
-; RUN: llc -mtriple=riscv32 -mattr=+v -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=2 | FileCheck %s --check-prefixes=NO_FOLDING
-; RUN: llc -mtriple=riscv64 -mattr=+v -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=2 | FileCheck %s --check-prefixes=NO_FOLDING
+; RUN: llc -mtriple=riscv32 -mattr=+v -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=1 | FileCheck %s --check-prefixes=NO_FOLDING1
+; RUN: llc -mtriple=riscv64 -mattr=+v -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=1 | FileCheck %s --check-prefixes=NO_FOLDING1
+; RUN: llc -mtriple=riscv32 -mattr=+v -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=2 | FileCheck %s --check-prefixes=NO_FOLDING2
+; RUN: llc -mtriple=riscv64 -mattr=+v -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=2 | FileCheck %s --check-prefixes=NO_FOLDING2
 ; RUN: llc -mtriple=riscv32 -mattr=+v -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=3 | FileCheck %s --check-prefixes=FOLDING
 ; RUN: llc -mtriple=riscv64 -mattr=+v -verify-machineinstrs %s -o - --riscv-lower-ext-max-web-size=3 | FileCheck %s --check-prefixes=FOLDING
 ; Check that the default value enables the web folding and
@@ -16,21 +16,38 @@
 ; We need the web size to be at least 3 for the folding to happen, because
 ; %c has 3 uses.
 define <2 x i16> @vwmul_v2i16_multiple_users(ptr %x, ptr %y, ptr %z) {
-; NO_FOLDING-LABEL: vwmul_v2i16_multiple_users:
-; NO_FOLDING:       # %bb.0:
-; NO_FOLDING-NEXT:    vsetivli zero, 2, e16, mf4, ta, ma
-; NO_FOLDING-NEXT:    vle8.v v8, (a0)
-; NO_FOLDING-NEXT:    vle8.v v9, (a1)
-; NO_FOLDING-NEXT:    vle8.v v10, (a2)
-; NO_FOLDING-NEXT:    vsext.vf2 v11, v8
-; NO_FOLDING-NEXT:    vsext.vf2 v8, v9
-; NO_FOLDING-NEXT:    vsext.vf2 v9, v10
-; NO_FOLDING-NEXT:    vmul.vv v8, v11, v8
-; NO_FOLDING-NEXT:    vadd.vv v10, v11, v9
-; NO_FOLDING-NEXT:    vsub.vv v9, v11, v9
-; NO_FOLDING-NEXT:    vor.vv v8, v8, v10
-; NO_FOLDING-NEXT:    vor.vv v8, v8, v9
-; NO_FOLDING-NEXT:    ret
+; NO_FOLDING1-LABEL: vwmul_v2i16_multiple_users:
+; NO_FOLDING1:       # %bb.0:
+; NO_FOLDING1-NEXT:    vsetivli zero, 2, e16, mf4, ta, ma
+; NO_FOLDING1-NEXT:    vle8.v v8, (a0)
+; NO_FOLDING1-NEXT:    vle8.v v9, (a1)
+; NO_FOLDING1-NEXT:    vle8.v v10, (a2)
+; NO_FOLDING1-NEXT:    vsext.vf2 v11, v8
+; NO_FOLDING1-NEXT:    vsext.vf2 v8, v9
+; NO_FOLDING1-NEXT:    vsext.vf2 v9, v10
+; NO_FOLDING1-NEXT:    vmul.vv v8, v11, v8
+; NO_FOLDING1-NEXT:    vadd.vv v10, v11, v9
+; NO_FOLDING1-NEXT:    vsub.vv v9, v11, v9
+; NO_FOLDING1-NEXT:    vor.vv v8, v8, v10
+; NO_FOLDING1-NEXT:    vor.vv v8, v8, v9
+; NO_FOLDING1-NEXT:    ret
+;
+; NO_FOLDING2-LABEL: vwmul_v2i16_multiple_users:
+; NO_FOLDING2:       # %bb.0:
+; NO_FOLDING2-NEXT:    vsetivli zero, 2, e16, mf4, ta, ma
+; NO_FOLDING2-NEXT:    vle8.v v8, (a0)
+; NO_FOLDING2-NEXT:    vle8.v v9, (a1)
+; NO_FOLDING2-NEXT:    vle8.v v10, (a2)
+; NO_FOLDING2-NEXT:    vsext.vf2 v11, v8
+; NO_FOLDING2-NEXT:    vsext.vf2 v8, v9
+; NO_FOLDING2-NEXT:    vmul.vv v8, v11, v8
+; NO_FOLDING2-NEXT:    vsetvli zero, zero, e8, mf8, ta, ma
+; NO_FOLDING2-NEXT:    vwadd.wv v9, v11, v10
+; NO_FOLDING2-NEXT:    vwsub.wv v11, v11, v10
+; NO_FOLDING2-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; NO_FOLDING2-NEXT:    vor.vv v8, v8, v9
+; NO_FOLDING2-NEXT:    vor.vv v8, v8, v11
+; NO_FOLDING2-NEXT:    ret
 ;
 ; FOLDING-LABEL: vwmul_v2i16_multiple_users:
 ; FOLDING:       # %bb.0:

llvm/test/CodeGen/X86/vector-shuffle-combining.ll

@@ -3572,45 +3572,53 @@ define void @SpinningCube() {
 ; SSE2-LABEL: SpinningCube:
 ; SSE2:       # %bb.0: # %entry
 ; SSE2-NEXT:    movl $1065353216, (%rax) # imm = 0x3F800000
-; SSE2-NEXT:    movaps {{.*#+}} xmm0 = [u,u,u,1.0E+0]
-; SSE2-NEXT:    movss {{.*#+}} xmm1 = [NaN,0.0E+0,0.0E+0,0.0E+0]
-; SSE2-NEXT:    movapd {{.*#+}} xmm2 = [u,u,-2.0E+0,u]
-; SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
-; SSE2-NEXT:    xorps %xmm3, %xmm3
-; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,3]
-; SSE2-NEXT:    addps %xmm3, %xmm1
-; SSE2-NEXT:    movaps %xmm1, (%rax)
-; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE2-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
-; SSE2-NEXT:    addps %xmm0, %xmm1
-; SSE2-NEXT:    movaps %xmm1, (%rax)
+; SSE2-NEXT:    xorps %xmm0, %xmm0
+; SSE2-NEXT:    movss {{.*#+}} xmm1 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,1]
+; SSE2-NEXT:    xorps %xmm2, %xmm2
+; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE2-NEXT:    movss {{.*#+}} xmm3 = [NaN,0.0E+0,0.0E+0,0.0E+0]
+; SSE2-NEXT:    movapd {{.*#+}} xmm4 = [u,u,-2.0E+0,u]
+; SSE2-NEXT:    movsd {{.*#+}} xmm4 = xmm3[0],xmm4[1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,0]
+; SSE2-NEXT:    movq {{.*#+}} xmm3 = xmm3[0],zero
+; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[2,0]
+; SSE2-NEXT:    addps %xmm0, %xmm3
+; SSE2-NEXT:    movaps %xmm3, (%rax)
+; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; SSE2-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; SSE2-NEXT:    addps %xmm2, %xmm0
+; SSE2-NEXT:    movaps %xmm0, (%rax)
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: SpinningCube:
 ; SSSE3:       # %bb.0: # %entry
 ; SSSE3-NEXT:    movl $1065353216, (%rax) # imm = 0x3F800000
-; SSSE3-NEXT:    movaps {{.*#+}} xmm0 = [u,u,u,1.0E+0]
-; SSSE3-NEXT:    movss {{.*#+}} xmm1 = [NaN,0.0E+0,0.0E+0,0.0E+0]
-; SSSE3-NEXT:    movapd {{.*#+}} xmm2 = [u,u,-2.0E+0,u]
-; SSSE3-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
-; SSSE3-NEXT:    xorps %xmm3, %xmm3
-; SSSE3-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,3]
-; SSSE3-NEXT:    addps %xmm3, %xmm1
-; SSSE3-NEXT:    movaps %xmm1, (%rax)
-; SSSE3-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0,0,2]
-; SSSE3-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
-; SSSE3-NEXT:    addps %xmm0, %xmm1
-; SSSE3-NEXT:    movaps %xmm1, (%rax)
+; SSSE3-NEXT:    xorps %xmm0, %xmm0
+; SSSE3-NEXT:    movss {{.*#+}} xmm1 = [1.0E+0,0.0E+0,0.0E+0,0.0E+0]
+; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,1]
+; SSSE3-NEXT:    xorps %xmm2, %xmm2
+; SSSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSSE3-NEXT:    movss {{.*#+}} xmm3 = [NaN,0.0E+0,0.0E+0,0.0E+0]
+; SSSE3-NEXT:    movapd {{.*#+}} xmm4 = [u,u,-2.0E+0,u]
+; SSSE3-NEXT:    movsd {{.*#+}} xmm4 = xmm3[0],xmm4[1]
+; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,0]
+; SSSE3-NEXT:    movq {{.*#+}} xmm3 = xmm3[0],zero
+; SSSE3-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[2,0]
+; SSSE3-NEXT:    addps %xmm0, %xmm3
+; SSSE3-NEXT:    movaps %xmm3, (%rax)
+; SSSE3-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; SSSE3-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; SSSE3-NEXT:    addps %xmm2, %xmm0
+; SSSE3-NEXT:    movaps %xmm0, (%rax)
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: SpinningCube:
 ; SSE41:       # %bb.0: # %entry
 ; SSE41-NEXT:    movl $1065353216, (%rax) # imm = 0x3F800000
-; SSE41-NEXT:    movaps {{.*#+}} xmm0 = [u,u,u,1.0E+0]
+; SSE41-NEXT:    insertps {{.*#+}} xmm0 = zero,zero,zero,mem[0]
 ; SSE41-NEXT:    movaps {{.*#+}} xmm1 = [0.0E+0,0.0E+0,-2.0E+0,u]
 ; SSE41-NEXT:    movss {{.*#+}} xmm2 = [NaN,0.0E+0,0.0E+0,0.0E+0]
 ; SSE41-NEXT:    movaps %xmm1, %xmm3
@@ -3629,7 +3637,7 @@ define void @SpinningCube() {
 ; AVX-LABEL: SpinningCube:
 ; AVX:       # %bb.0: # %entry
 ; AVX-NEXT:    movl $1065353216, (%rax) # imm = 0x3F800000
-; AVX-NEXT:    vbroadcastss {{.*#+}} xmm0 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
+; AVX-NEXT:    vmovaps {{.*#+}} xmm0 = [0.0E+0,0.0E+0,0.0E+0,1.0E+0]
 ; AVX-NEXT:    vmovaps {{.*#+}} xmm1 = [0.0E+0,0.0E+0,-2.0E+0,u]
 ; AVX-NEXT:    vmovss {{.*#+}} xmm2 = [NaN,0.0E+0,0.0E+0,0.0E+0]
 ; AVX-NEXT:    vinsertps {{.*#+}} xmm3 = xmm1[0,1,2],xmm2[0]