shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity
llvm-project/llvm/test/Transforms/LoopVectorize/iv_outside_user.ll
Florian Hahn b20b6e9ea9
[LV] Check IR generated for both interleaving and vectorizing in test. 
Currently the tests would in some cases would only check the vectorized
IR, but not the interleaved IR, if they are different.
2024-12-30 19:57:26 +00:00

982 lines
52 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --replace-value-regex "!llvm.loop ![0-9]+" --version 5
; RUN: opt -S -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=2 < %s | FileCheck --check-prefixes=CHECK,VEC %s
; RUN: opt -S -passes=loop-vectorize -force-vector-interleave=2 -force-vector-width=1 < %s | FileCheck --check-prefixes=CHECK,INTERLEAVE %s
define i32 @postinc(i32 %k) {
; CHECK-LABEL: define i32 @postinc(
; CHECK-SAME: i32 [[K:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[K]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP0]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[FOR_BODY:.*]]
; CHECK: [[FOR_BODY]]:
; CHECK-NEXT: [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[INC]] = add nsw i32 [[INC_PHI]], 1
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[INC]], [[K]]
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[FOR_END]]:
; CHECK-NEXT: [[INC_LCSSA:%.*]] = phi i32 [ [[INC]], %[[FOR_BODY]] ], [ [[N_VEC]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[INC_LCSSA]]
;
entry:
br label %for.body
for.body:
%inc.phi = phi i32 [ 0, %entry ], [ %inc, %for.body ]
%inc = add nsw i32 %inc.phi, 1
%cmp = icmp eq i32 %inc, %k
br i1 %cmp, label %for.end, label %for.body
for.end:
ret i32 %inc
}
define i32 @preinc(i32 %k) {
; CHECK-LABEL: define i32 @preinc(
; CHECK-SAME: i32 [[K:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[K]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP0]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]
; CHECK-NEXT: [[IND_ESCAPE:%.*]] = sub i32 [[N_VEC]], 1
; CHECK-NEXT: br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[FOR_BODY:.*]]
; CHECK: [[FOR_BODY]]:
; CHECK-NEXT: [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[INC]] = add nsw i32 [[INC_PHI]], 1
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[INC]], [[K]]
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[FOR_END]]:
; CHECK-NEXT: [[INC_PHI_LCSSA:%.*]] = phi i32 [ [[INC_PHI]], %[[FOR_BODY]] ], [ [[IND_ESCAPE]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[INC_PHI_LCSSA]]
;
entry:
br label %for.body
for.body:
%inc.phi = phi i32 [ 0, %entry ], [ %inc, %for.body ]
%inc = add nsw i32 %inc.phi, 1
%cmp = icmp eq i32 %inc, %k
br i1 %cmp, label %for.end, label %for.body
for.end:
ret i32 %inc.phi
}
define i32 @constpre() {
; CHECK-LABEL: define i32 @constpre() {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
; CHECK-NEXT: br i1 [[TMP0]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: br i1 true, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[MIDDLE_BLOCK]] ], [ 32, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[FOR_BODY:.*]]
; CHECK: [[FOR_BODY]]:
; CHECK-NEXT: [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[INC]] = sub nsw i32 [[INC_PHI]], 2
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[INC]], 0
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[FOR_END]]:
; CHECK-NEXT: [[INC_PHI_LCSSA:%.*]] = phi i32 [ [[INC_PHI]], %[[FOR_BODY]] ], [ 2, %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[INC_PHI_LCSSA]]
;
entry:
br label %for.body
for.body:
%inc.phi = phi i32 [ 32, %entry ], [ %inc, %for.body ]
%inc = sub nsw i32 %inc.phi, 2
%cmp = icmp eq i32 %inc, 0
br i1 %cmp, label %for.end, label %for.body
for.end:
ret i32 %inc.phi
}
define ptr @geppre(ptr %ptr) {
; CHECK-LABEL: define ptr @geppre(
; CHECK-SAME: ptr [[PTR:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[IND_END:%.*]] = getelementptr i8, ptr [[PTR]], i64 512
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[TMP0:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
; CHECK-NEXT: br i1 [[TMP0]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[IND_ESCAPE:%.*]] = getelementptr i8, ptr [[IND_END]], i64 -16
; CHECK-NEXT: br i1 true, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 32, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[PTR]], %[[ENTRY]] ]
; CHECK-NEXT: br label %[[FOR_BODY:.*]]
; CHECK: [[FOR_BODY]]:
; CHECK-NEXT: [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[PTR_PHI:%.*]] = phi ptr [ [[BC_RESUME_VAL1]], %[[SCALAR_PH]] ], [ [[INC_PTR:%.*]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[INC]] = add nsw i32 [[INC_PHI]], 1
; CHECK-NEXT: [[INC_PTR]] = getelementptr i32, ptr [[PTR_PHI]], i32 4
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[INC]], 32
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[FOR_END]]:
; CHECK-NEXT: [[PTR_PHI_LCSSA:%.*]] = phi ptr [ [[PTR_PHI]], %[[FOR_BODY]] ], [ [[IND_ESCAPE]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret ptr [[PTR_PHI_LCSSA]]
;
entry:
br label %for.body
for.body:
%inc.phi = phi i32 [ 0, %entry ], [ %inc, %for.body ]
%ptr.phi = phi ptr [ %ptr, %entry ], [ %inc.ptr, %for.body ]
%inc = add nsw i32 %inc.phi, 1
%inc.ptr = getelementptr i32, ptr %ptr.phi, i32 4
%cmp = icmp eq i32 %inc, 32
br i1 %cmp, label %for.end, label %for.body
for.end:
ret ptr %ptr.phi
}
define ptr @both(i32 %k) {
; CHECK-LABEL: define ptr @both(
; CHECK-SAME: i32 [[K:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[BASE:%.*]] = getelementptr inbounds i32, ptr undef, i64 1
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[K]], -1
; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; CHECK-NEXT: [[IND_END:%.*]] = trunc i64 [[N_VEC]] to i32
; CHECK-NEXT: [[TMP3:%.*]] = mul i64 [[N_VEC]], 4
; CHECK-NEXT: [[IND_END1:%.*]] = getelementptr i8, ptr [[BASE]], i64 [[TMP3]]
; CHECK-NEXT: [[TMP4:%.*]] = mul i64 [[N_VEC]], 4
; CHECK-NEXT: [[IND_END2:%.*]] = getelementptr i8, ptr undef, i64 [[TMP4]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; CHECK-NEXT: [[IND_ESCAPE:%.*]] = getelementptr i8, ptr [[IND_END1]], i64 -4
; CHECK-NEXT: br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_RESUME_VAL2:%.*]] = phi ptr [ [[IND_END1]], %[[MIDDLE_BLOCK]] ], [ [[BASE]], %[[ENTRY]] ]
; CHECK-NEXT: [[BC_RESUME_VAL4:%.*]] = phi ptr [ [[IND_END2]], %[[MIDDLE_BLOCK]] ], [ undef, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[FOR_BODY:.*]]
; CHECK: [[FOR_BODY]]:
; CHECK-NEXT: [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[INC_LAG1:%.*]] = phi ptr [ [[BC_RESUME_VAL2]], %[[SCALAR_PH]] ], [ [[TMP:%.*]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[INC_LAG2:%.*]] = phi ptr [ [[BC_RESUME_VAL4]], %[[SCALAR_PH]] ], [ [[INC_LAG1]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[TMP]] = getelementptr inbounds i32, ptr [[INC_LAG1]], i64 1
; CHECK-NEXT: [[INC]] = add nsw i32 [[INC_PHI]], 1
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[INC]], [[K]]
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[FOR_END]]:
; CHECK-NEXT: [[INC_LAG1_LCSSA:%.*]] = phi ptr [ [[INC_LAG1]], %[[FOR_BODY]] ], [ [[IND_ESCAPE]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret ptr [[INC_LAG1_LCSSA]]
;
entry:
%base = getelementptr inbounds i32, ptr undef, i64 1
br label %for.body
for.body:
%inc.phi = phi i32 [ 0, %entry ], [ %inc, %for.body ]
%inc.lag1 = phi ptr [ %base, %entry ], [ %tmp, %for.body]
%inc.lag2 = phi ptr [ undef, %entry ], [ %inc.lag1, %for.body]
%tmp = getelementptr inbounds i32, ptr %inc.lag1, i64 1
%inc = add nsw i32 %inc.phi, 1
%cmp = icmp eq i32 %inc, %k
br i1 %cmp, label %for.end, label %for.body
for.end:
ret ptr %inc.lag1
}
define i32 @multiphi(i32 %k, ptr %p) {
; CHECK-LABEL: define i32 @multiphi(
; CHECK-SAME: i32 [[K:%.*]], ptr [[P:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[K]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP0]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[FOR_BODY:.*]]
; CHECK: [[FOR_BODY]]:
; CHECK-NEXT: [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[INC]] = add nsw i32 [[INC_PHI]], 1
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[INC]], [[K]]
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[FOR_END]]:
; CHECK-NEXT: [[PHI:%.*]] = phi i32 [ [[INC]], %[[FOR_BODY]] ], [ [[N_VEC]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: [[PHI2:%.*]] = phi i32 [ [[INC]], %[[FOR_BODY]] ], [ [[N_VEC]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: store i32 [[PHI2]], ptr [[P]], align 4
; CHECK-NEXT: ret i32 [[PHI]]
;
entry:
br label %for.body
for.body:
%inc.phi = phi i32 [ 0, %entry ], [ %inc, %for.body ]
%inc = add nsw i32 %inc.phi, 1
%cmp = icmp eq i32 %inc, %k
br i1 %cmp, label %for.end, label %for.body
for.end:
%phi = phi i32 [ %inc, %for.body ]
%phi2 = phi i32 [ %inc, %for.body ]
store i32 %phi2, ptr %p
ret i32 %phi
}
define void @PR30742() {
; CHECK-LABEL: define void @PR30742() {
; CHECK-NEXT: [[BB0:.*:]]
; CHECK-NEXT: br label %[[BB1:.*]]
; CHECK: [[BB1_LOOPEXIT:.*]]:
; CHECK-NEXT: br label %[[BB1]]
; CHECK: [[BB1]]:
; CHECK-NEXT: [[TMP00:%.*]] = load i32, ptr undef, align 16
; CHECK-NEXT: [[TMP01:%.*]] = sub i32 [[TMP00]], undef
; CHECK-NEXT: [[TMP02:%.*]] = icmp slt i32 [[TMP01]], 1
; CHECK-NEXT: [[TMP03:%.*]] = select i1 [[TMP02]], i32 1, i32 [[TMP01]]
; CHECK-NEXT: [[TMP04:%.*]] = add nsw i32 [[TMP03]], -7
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[TMP03]], -8
; CHECK-NEXT: [[TMP1:%.*]] = add nsw i32 [[TMP03]], -15
; CHECK-NEXT: [[SMIN1:%.*]] = call i32 @llvm.smin.i32(i32 [[TMP1]], i32 0)
; CHECK-NEXT: [[TMP2:%.*]] = sub i32 [[TMP0]], [[SMIN1]]
; CHECK-NEXT: [[TMP3:%.*]] = lshr i32 [[TMP2]], 3
; CHECK-NEXT: [[TMP4:%.*]] = add nuw nsw i32 [[TMP3]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK4:%.*]] = icmp ult i32 [[TMP4]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK4]], label %[[SCALAR_PH3:.*]], label %[[VECTOR_PH5:.*]]
; CHECK: [[VECTOR_PH5]]:
; CHECK-NEXT: [[N_MOD_VF6:%.*]] = urem i32 [[TMP4]], 2
; CHECK-NEXT: [[N_VEC7:%.*]] = sub i32 [[TMP4]], [[N_MOD_VF6]]
; CHECK-NEXT: [[TMP5:%.*]] = mul i32 [[N_VEC7]], -8
; CHECK-NEXT: [[IND_END8:%.*]] = add i32 [[TMP04]], [[TMP5]]
; CHECK-NEXT: br label %[[VECTOR_BODY10:.*]]
; CHECK: [[VECTOR_BODY10]]:
; CHECK-NEXT: [[INDEX10:%.*]] = phi i32 [ 0, %[[VECTOR_PH5]] ], [ [[INDEX_NEXT11:%.*]], %[[VECTOR_BODY10]] ]
; CHECK-NEXT: [[INDEX_NEXT11]] = add nuw i32 [[INDEX10]], 2
; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT11]], [[N_VEC7]]
; CHECK-NEXT: br i1 [[TMP6]], label %[[MIDDLE_BLOCK2:.*]], label %[[VECTOR_BODY10]], {{!llvm.loop ![0-9]+}}
; CHECK: [[MIDDLE_BLOCK2]]:
; CHECK-NEXT: [[CMP_N12:%.*]] = icmp eq i32 [[TMP4]], [[N_VEC7]]
; CHECK-NEXT: [[IND_ESCAPE:%.*]] = sub i32 [[IND_END8]], -8
; CHECK-NEXT: br i1 [[CMP_N12]], label %[[BB3:.*]], label %[[SCALAR_PH3]]
; CHECK: [[SCALAR_PH3]]:
; CHECK-NEXT: [[BC_RESUME_VAL9:%.*]] = phi i32 [ [[IND_END8]], %[[MIDDLE_BLOCK2]] ], [ [[TMP04]], %[[BB1]] ]
; CHECK-NEXT: br label %[[BB2:.*]]
; CHECK: [[BB2]]:
; CHECK-NEXT: [[TMP05:%.*]] = phi i32 [ [[BC_RESUME_VAL9]], %[[SCALAR_PH3]] ], [ [[TMP06:%.*]], %[[BB2]] ]
; CHECK-NEXT: [[TMP06]] = add i32 [[TMP05]], -8
; CHECK-NEXT: [[TMP07:%.*]] = icmp sgt i32 [[TMP06]], 0
; CHECK-NEXT: br i1 [[TMP07]], label %[[BB2]], label %[[BB3]], {{!llvm.loop ![0-9]+}}
; CHECK: [[BB3]]:
; CHECK-NEXT: [[TMP08:%.*]] = phi i32 [ [[TMP05]], %[[BB2]] ], [ [[IND_ESCAPE]], %[[MIDDLE_BLOCK2]] ]
; CHECK-NEXT: [[TMP09:%.*]] = sub i32 [[TMP00]], undef
; CHECK-NEXT: [[TMP10:%.*]] = icmp slt i32 [[TMP09]], 1
; CHECK-NEXT: [[TMP11:%.*]] = select i1 [[TMP10]], i32 1, i32 [[TMP09]]
; CHECK-NEXT: [[TMP12:%.*]] = add nsw i32 [[TMP11]], -7
; CHECK-NEXT: [[TMP7:%.*]] = add i32 [[TMP11]], -8
; CHECK-NEXT: [[TMP8:%.*]] = add nsw i32 [[TMP11]], -15
; CHECK-NEXT: [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[TMP8]], i32 0)
; CHECK-NEXT: [[TMP9:%.*]] = sub i32 [[TMP7]], [[SMIN]]
; CHECK-NEXT: [[TMP10:%.*]] = lshr i32 [[TMP9]], 3
; CHECK-NEXT: [[TMP11:%.*]] = add nuw nsw i32 [[TMP10]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP11]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP11]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP11]], [[N_MOD_VF]]
; CHECK-NEXT: [[TMP16:%.*]] = mul i32 [[N_VEC]], -8
; CHECK-NEXT: [[IND_END:%.*]] = add i32 [[TMP12]], [[TMP16]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP11]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label %[[BB1_LOOPEXIT]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[TMP12]], %[[BB3]] ]
; CHECK-NEXT: br label %[[BB4:.*]]
; CHECK: [[BB4]]:
; CHECK-NEXT: [[TMP13:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[TMP14:%.*]], %[[BB4]] ]
; CHECK-NEXT: [[TMP14]] = add i32 [[TMP13]], -8
; CHECK-NEXT: [[TMP15:%.*]] = icmp sgt i32 [[TMP14]], 0
; CHECK-NEXT: br i1 [[TMP15]], label %[[BB4]], label %[[BB1_LOOPEXIT]], {{!llvm.loop ![0-9]+}}
;
BB0:
br label %BB1
BB1:
%tmp00 = load i32, ptr undef, align 16
%tmp01 = sub i32 %tmp00, undef
%tmp02 = icmp slt i32 %tmp01, 1
%tmp03 = select i1 %tmp02, i32 1, i32 %tmp01
%tmp04 = add nsw i32 %tmp03, -7
br label %BB2
BB2:
%tmp05 = phi i32 [ %tmp04, %BB1 ], [ %tmp06, %BB2 ]
%tmp06 = add i32 %tmp05, -8
%tmp07 = icmp sgt i32 %tmp06, 0
br i1 %tmp07, label %BB2, label %BB3
BB3:
%tmp08 = phi i32 [ %tmp05, %BB2 ]
%tmp09 = sub i32 %tmp00, undef
%tmp10 = icmp slt i32 %tmp09, 1
%tmp11 = select i1 %tmp10, i32 1, i32 %tmp09
%tmp11.inc = add nsw i32 %tmp11, -7
br label %BB4
BB4:
%tmp13 = phi i32 [ %tmp11.inc, %BB3 ], [ %tmp14, %BB4 ]
%tmp14 = add i32 %tmp13, -8
%tmp15 = icmp sgt i32 %tmp14, 0
br i1 %tmp15, label %BB4, label %BB1
}
;
define i64 @iv_scalar_steps_and_outside_users(ptr %ptr) {
; VEC-LABEL: define i64 @iv_scalar_steps_and_outside_users(
; VEC-SAME: ptr [[PTR:%.*]]) {
; VEC-NEXT: [[ENTRY:.*]]:
; VEC-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; VEC: [[VECTOR_PH]]:
; VEC-NEXT: br label %[[VECTOR_BODY:.*]]
; VEC: [[VECTOR_BODY]]:
; VEC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VEC-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VEC-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
; VEC-NEXT: [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[TMP0]]
; VEC-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
; VEC-NEXT: store <2 x i64> [[VEC_IND]], ptr [[TMP2]], align 4
; VEC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; VEC-NEXT: [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
; VEC-NEXT: [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1002
; VEC-NEXT: br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; VEC: [[MIDDLE_BLOCK]]:
; VEC-NEXT: br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
; VEC: [[SCALAR_PH]]:
; VEC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1002, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; VEC-NEXT: br label %[[LOOP:.*]]
; VEC: [[LOOP]]:
; VEC-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; VEC-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 1
; VEC-NEXT: [[GEP_PTR:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[IV]]
; VEC-NEXT: store i64 [[IV]], ptr [[GEP_PTR]], align 4
; VEC-NEXT: [[EXITCOND:%.*]] = icmp ugt i64 [[IV]], 1000
; VEC-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], {{!llvm.loop ![0-9]+}}
; VEC: [[EXIT]]:
; VEC-NEXT: [[IV_LCSSA:%.*]] = phi i64 [ [[IV]], %[[LOOP]] ], [ 1001, %[[MIDDLE_BLOCK]] ]
; VEC-NEXT: ret i64 [[IV_LCSSA]]
;
; INTERLEAVE-LABEL: define i64 @iv_scalar_steps_and_outside_users(
; INTERLEAVE-SAME: ptr [[PTR:%.*]]) {
; INTERLEAVE-NEXT: [[ENTRY:.*]]:
; INTERLEAVE-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; INTERLEAVE: [[VECTOR_PH]]:
; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]]
; INTERLEAVE: [[VECTOR_BODY]]:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
; INTERLEAVE-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[TMP0]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[TMP1]]
; INTERLEAVE-NEXT: store i64 [[TMP0]], ptr [[TMP2]], align 4
; INTERLEAVE-NEXT: store i64 [[TMP1]], ptr [[TMP3]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; INTERLEAVE-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1002
; INTERLEAVE-NEXT: br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; INTERLEAVE: [[MIDDLE_BLOCK]]:
; INTERLEAVE-NEXT: br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
; INTERLEAVE: [[SCALAR_PH]]:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1002, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: br label %[[LOOP:.*]]
; INTERLEAVE: [[LOOP]]:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 1
; INTERLEAVE-NEXT: [[GEP_PTR:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[IV]]
; INTERLEAVE-NEXT: store i64 [[IV]], ptr [[GEP_PTR]], align 4
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp ugt i64 [[IV]], 1000
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], {{!llvm.loop ![0-9]+}}
; INTERLEAVE: [[EXIT]]:
; INTERLEAVE-NEXT: [[IV_LCSSA:%.*]] = phi i64 [ [[IV]], %[[LOOP]] ], [ 1001, %[[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: ret i64 [[IV_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%iv.next = add nuw i64 %iv, 1
%gep.ptr = getelementptr inbounds i64, ptr %ptr, i64 %iv
store i64 %iv, ptr %gep.ptr
%exitcond = icmp ugt i64 %iv, 1000
br i1 %exitcond, label %exit, label %loop
exit:
%iv.lcssa = phi i64 [ %iv, %loop ]
ret i64 %iv.lcssa
}
; %iv.2 is dead in the vector loop and only used outside the loop.
define i32 @iv_2_dead_in_loop_only_used_outside(ptr %ptr) {
; VEC-LABEL: define i32 @iv_2_dead_in_loop_only_used_outside(
; VEC-SAME: ptr [[PTR:%.*]]) {
; VEC-NEXT: [[ENTRY:.*]]:
; VEC-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; VEC: [[VECTOR_PH]]:
; VEC-NEXT: br label %[[VECTOR_BODY:.*]]
; VEC: [[VECTOR_BODY]]:
; VEC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VEC-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VEC-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
; VEC-NEXT: [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[TMP0]]
; VEC-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
; VEC-NEXT: store <2 x i64> [[VEC_IND]], ptr [[TMP2]], align 4
; VEC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; VEC-NEXT: [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
; VEC-NEXT: [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1002
; VEC-NEXT: br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; VEC: [[MIDDLE_BLOCK]]:
; VEC-NEXT: br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
; VEC: [[SCALAR_PH]]:
; VEC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1002, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; VEC-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 2004, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; VEC-NEXT: br label %[[LOOP:.*]]
; VEC: [[LOOP]]:
; VEC-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; VEC-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], %[[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], %[[LOOP]] ]
; VEC-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 1
; VEC-NEXT: [[IV_2_NEXT]] = add nuw i32 [[IV_2]], 2
; VEC-NEXT: [[GEP_PTR:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[IV]]
; VEC-NEXT: store i64 [[IV]], ptr [[GEP_PTR]], align 4
; VEC-NEXT: [[EXITCOND:%.*]] = icmp ugt i64 [[IV]], 1000
; VEC-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], {{!llvm.loop ![0-9]+}}
; VEC: [[EXIT]]:
; VEC-NEXT: [[IV_2_LCSSA:%.*]] = phi i32 [ [[IV_2]], %[[LOOP]] ], [ 2002, %[[MIDDLE_BLOCK]] ]
; VEC-NEXT: ret i32 [[IV_2_LCSSA]]
;
; INTERLEAVE-LABEL: define i32 @iv_2_dead_in_loop_only_used_outside(
; INTERLEAVE-SAME: ptr [[PTR:%.*]]) {
; INTERLEAVE-NEXT: [[ENTRY:.*]]:
; INTERLEAVE-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; INTERLEAVE: [[VECTOR_PH]]:
; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]]
; INTERLEAVE: [[VECTOR_BODY]]:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
; INTERLEAVE-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[TMP0]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[TMP1]]
; INTERLEAVE-NEXT: store i64 [[TMP0]], ptr [[TMP2]], align 4
; INTERLEAVE-NEXT: store i64 [[TMP1]], ptr [[TMP3]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; INTERLEAVE-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1002
; INTERLEAVE-NEXT: br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; INTERLEAVE: [[MIDDLE_BLOCK]]:
; INTERLEAVE-NEXT: br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
; INTERLEAVE: [[SCALAR_PH]]:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1002, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 2004, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: br label %[[LOOP:.*]]
; INTERLEAVE: [[LOOP]]:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], %[[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 1
; INTERLEAVE-NEXT: [[IV_2_NEXT]] = add nuw i32 [[IV_2]], 2
; INTERLEAVE-NEXT: [[GEP_PTR:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[IV]]
; INTERLEAVE-NEXT: store i64 [[IV]], ptr [[GEP_PTR]], align 4
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp ugt i64 [[IV]], 1000
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], {{!llvm.loop ![0-9]+}}
; INTERLEAVE: [[EXIT]]:
; INTERLEAVE-NEXT: [[IV_2_LCSSA:%.*]] = phi i32 [ [[IV_2]], %[[LOOP]] ], [ 2002, %[[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: ret i32 [[IV_2_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%iv.2 = phi i32 [ 0, %entry ], [ %iv.2.next, %loop ]
%iv.next = add nuw i64 %iv, 1
%iv.2.next = add nuw i32 %iv.2, 2
%gep.ptr = getelementptr inbounds i64, ptr %ptr, i64 %iv
store i64 %iv, ptr %gep.ptr
%exitcond = icmp ugt i64 %iv, 1000
br i1 %exitcond, label %exit, label %loop
exit:
%iv.2.lcssa = phi i32 [ %iv.2, %loop ]
ret i32 %iv.2.lcssa
}
define i32 @postinc_sub(i32 %k) {
; CHECK-LABEL: define i32 @postinc_sub(
; CHECK-SAME: i32 [[K:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[K]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]
; CHECK-NEXT: [[TMP0:%.*]] = sub i32 [[K]], [[N_VEC]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP1]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ [[K]], %[[ENTRY]] ]
; CHECK-NEXT: br label %[[FOR_BODY:.*]]
; CHECK: [[FOR_BODY]]:
; CHECK-NEXT: [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[INC]] = sub nsw i32 [[INC_PHI]], 1
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[INC]], 0
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[FOR_END]]:
; CHECK-NEXT: [[INC_LCSSA:%.*]] = phi i32 [ [[INC]], %[[FOR_BODY]] ], [ [[TMP0]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[INC_LCSSA]]
;
entry:
br label %for.body
for.body:
%inc.phi = phi i32 [ %k, %entry ], [ %inc, %for.body ]
%inc = sub nsw i32 %inc.phi, 1
%cmp = icmp eq i32 %inc, 0
br i1 %cmp, label %for.end, label %for.body
for.end:
ret i32 %inc
}
define i32 @postinc_swapped_ops(i32 %k) {
; CHECK-LABEL: define i32 @postinc_swapped_ops(
; CHECK-SAME: i32 [[K:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[K]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP0]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[FOR_BODY:.*]]
; CHECK: [[FOR_BODY]]:
; CHECK-NEXT: [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
; CHECK-NEXT: [[INC]] = add nsw i32 1, [[INC_PHI]]
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[INC]], [[K]]
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_BODY]], {{!llvm.loop ![0-9]+}}
; CHECK: [[FOR_END]]:
; CHECK-NEXT: [[INC_LCSSA:%.*]] = phi i32 [ [[INC]], %[[FOR_BODY]] ], [ [[N_VEC]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[INC_LCSSA]]
;
entry:
br label %for.body
for.body:
%inc.phi = phi i32 [ 0, %entry ], [ %inc, %for.body ]
%inc = add nsw i32 1, %inc.phi
%cmp = icmp eq i32 %inc, %k
br i1 %cmp, label %for.end, label %for.body
for.end:
ret i32 %inc
}
define i32 @postinc_not_iv_backedge_value(i32 %k) {
; VEC-LABEL: define i32 @postinc_not_iv_backedge_value(
; VEC-SAME: i32 [[K:%.*]]) {
; VEC-NEXT: [[ENTRY:.*]]:
; VEC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K]], 2
; VEC-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; VEC: [[VECTOR_PH]]:
; VEC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[K]], 2
; VEC-NEXT: [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]
; VEC-NEXT: br label %[[VECTOR_BODY:.*]]
; VEC: [[VECTOR_BODY]]:
; VEC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VEC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VEC-NEXT: [[TMP0:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
; VEC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; VEC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
; VEC-NEXT: [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; VEC-NEXT: br i1 [[TMP1]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; VEC: [[MIDDLE_BLOCK]]:
; VEC-NEXT: [[TMP2:%.*]] = extractelement <2 x i32> [[TMP0]], i32 1
; VEC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]
; VEC-NEXT: br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; VEC: [[SCALAR_PH]]:
; VEC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; VEC-NEXT: br label %[[FOR_BODY:.*]]
; VEC: [[FOR_BODY]]:
; VEC-NEXT: [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
; VEC-NEXT: [[INC]] = add nsw i32 [[INC_PHI]], 1
; VEC-NEXT: [[INC_2:%.*]] = add i32 [[INC_PHI]], 2
; VEC-NEXT: [[CMP:%.*]] = icmp eq i32 [[INC]], [[K]]
; VEC-NEXT: br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_BODY]], {{!llvm.loop ![0-9]+}}
; VEC: [[FOR_END]]:
; VEC-NEXT: [[INC_2_LCSSA:%.*]] = phi i32 [ [[INC_2]], %[[FOR_BODY]] ], [ [[TMP2]], %[[MIDDLE_BLOCK]] ]
; VEC-NEXT: ret i32 [[INC_2_LCSSA]]
;
; INTERLEAVE-LABEL: define i32 @postinc_not_iv_backedge_value(
; INTERLEAVE-SAME: i32 [[K:%.*]]) {
; INTERLEAVE-NEXT: [[ENTRY:.*]]:
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K]], 2
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; INTERLEAVE: [[VECTOR_PH]]:
; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[K]], 2
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]
; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]]
; INTERLEAVE: [[VECTOR_BODY]]:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 1
; INTERLEAVE-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], 2
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; INTERLEAVE-NEXT: [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP2]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; INTERLEAVE: [[MIDDLE_BLOCK]]:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
; INTERLEAVE: [[SCALAR_PH]]:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: br label %[[FOR_BODY:.*]]
; INTERLEAVE: [[FOR_BODY]]:
; INTERLEAVE-NEXT: [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
; INTERLEAVE-NEXT: [[INC]] = add nsw i32 [[INC_PHI]], 1
; INTERLEAVE-NEXT: [[INC_2:%.*]] = add i32 [[INC_PHI]], 2
; INTERLEAVE-NEXT: [[CMP:%.*]] = icmp eq i32 [[INC]], [[K]]
; INTERLEAVE-NEXT: br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_BODY]], {{!llvm.loop ![0-9]+}}
; INTERLEAVE: [[FOR_END]]:
; INTERLEAVE-NEXT: [[INC_2_LCSSA:%.*]] = phi i32 [ [[INC_2]], %[[FOR_BODY]] ], [ [[TMP1]], %[[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: ret i32 [[INC_2_LCSSA]]
;
entry:
br label %for.body
for.body:
%inc.phi = phi i32 [ 0, %entry ], [ %inc, %for.body ]
%inc = add nsw i32 %inc.phi, 1
%inc.2 = add i32 %inc.phi, 2
%cmp = icmp eq i32 %inc, %k
br i1 %cmp, label %for.end, label %for.body
for.end:
ret i32 %inc.2
}
define float @fp_postinc_use_fadd(float %init, ptr noalias nocapture %A, i64 %N, float %fpinc) {
; VEC-LABEL: define float @fp_postinc_use_fadd(
; VEC-SAME: float [[INIT:%.*]], ptr noalias nocapture [[A:%.*]], i64 [[N:%.*]], float [[FPINC:%.*]]) {
; VEC-NEXT: [[ENTRY:.*]]:
; VEC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 2
; VEC-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; VEC: [[VECTOR_PH]]:
; VEC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
; VEC-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; VEC-NEXT: [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
; VEC-NEXT: [[TMP0:%.*]] = fmul fast float [[FPINC]], [[DOTCAST]]
; VEC-NEXT: [[TMP1:%.*]] = fadd fast float [[INIT]], [[TMP0]]
; VEC-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[INIT]], i64 0
; VEC-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
; VEC-NEXT: [[DOTSPLATINSERT1:%.*]] = insertelement <2 x float> poison, float [[FPINC]], i64 0
; VEC-NEXT: [[DOTSPLAT2:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT1]], <2 x float> poison, <2 x i32> zeroinitializer
; VEC-NEXT: [[TMP2:%.*]] = fmul fast <2 x float> <float 0.000000e+00, float 1.000000e+00>, [[DOTSPLAT2]]
; VEC-NEXT: [[INDUCTION:%.*]] = fadd fast <2 x float> [[DOTSPLAT]], [[TMP2]]
; VEC-NEXT: [[TMP3:%.*]] = fmul fast float [[FPINC]], 2.000000e+00
; VEC-NEXT: [[DOTSPLATINSERT3:%.*]] = insertelement <2 x float> poison, float [[TMP3]], i64 0
; VEC-NEXT: [[DOTSPLAT4:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT3]], <2 x float> poison, <2 x i32> zeroinitializer
; VEC-NEXT: br label %[[VECTOR_BODY:.*]]
; VEC: [[VECTOR_BODY]]:
; VEC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VEC-NEXT: [[VEC_IND:%.*]] = phi <2 x float> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VEC-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 0
; VEC-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
; VEC-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
; VEC-NEXT: store <2 x float> [[VEC_IND]], ptr [[TMP6]], align 4
; VEC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; VEC-NEXT: [[VEC_IND_NEXT]] = fadd fast <2 x float> [[VEC_IND]], [[DOTSPLAT4]]
; VEC-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; VEC-NEXT: br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; VEC: [[MIDDLE_BLOCK]]:
; VEC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; VEC-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; VEC: [[SCALAR_PH]]:
; VEC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; VEC-NEXT: [[BC_RESUME_VAL5:%.*]] = phi float [ [[TMP1]], %[[MIDDLE_BLOCK]] ], [ [[INIT]], %[[ENTRY]] ]
; VEC-NEXT: br label %[[LOOP:.*]]
; VEC: [[LOOP]]:
; VEC-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; VEC-NEXT: [[FP_IV:%.*]] = phi float [ [[BC_RESUME_VAL5]], %[[SCALAR_PH]] ], [ [[ADD:%.*]], %[[LOOP]] ]
; VEC-NEXT: [[GEP_A:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
; VEC-NEXT: store float [[FP_IV]], ptr [[GEP_A]], align 4
; VEC-NEXT: [[ADD]] = fadd fast float [[FP_IV]], [[FPINC]]
; VEC-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; VEC-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; VEC-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], {{!llvm.loop ![0-9]+}}
; VEC: [[EXIT]]:
; VEC-NEXT: [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], %[[LOOP]] ], [ [[TMP1]], %[[MIDDLE_BLOCK]] ]
; VEC-NEXT: ret float [[ADD_LCSSA]]
;
; INTERLEAVE-LABEL: define float @fp_postinc_use_fadd(
; INTERLEAVE-SAME: float [[INIT:%.*]], ptr noalias nocapture [[A:%.*]], i64 [[N:%.*]], float [[FPINC:%.*]]) {
; INTERLEAVE-NEXT: [[ENTRY:.*]]:
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 2
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; INTERLEAVE: [[VECTOR_PH]]:
; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; INTERLEAVE-NEXT: [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
; INTERLEAVE-NEXT: [[TMP0:%.*]] = fmul fast float [[FPINC]], [[DOTCAST]]
; INTERLEAVE-NEXT: [[TMP1:%.*]] = fadd fast float [[INIT]], [[TMP0]]
; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]]
; INTERLEAVE: [[VECTOR_BODY]]:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 0
; INTERLEAVE-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 1
; INTERLEAVE-NEXT: [[DOTCAST1:%.*]] = sitofp i64 [[INDEX]] to float
; INTERLEAVE-NEXT: [[TMP4:%.*]] = fmul fast float [[FPINC]], [[DOTCAST1]]
; INTERLEAVE-NEXT: [[OFFSET_IDX:%.*]] = fadd fast float [[INIT]], [[TMP4]]
; INTERLEAVE-NEXT: [[TMP5:%.*]] = fmul fast float 0.000000e+00, [[FPINC]]
; INTERLEAVE-NEXT: [[TMP6:%.*]] = fadd fast float [[OFFSET_IDX]], [[TMP5]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = fmul fast float 1.000000e+00, [[FPINC]]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = fadd fast float [[OFFSET_IDX]], [[TMP7]]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP2]]
; INTERLEAVE-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP3]]
; INTERLEAVE-NEXT: store float [[TMP6]], ptr [[TMP9]], align 4
; INTERLEAVE-NEXT: store float [[TMP8]], ptr [[TMP10]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; INTERLEAVE-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; INTERLEAVE: [[MIDDLE_BLOCK]]:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; INTERLEAVE: [[SCALAR_PH]]:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL2:%.*]] = phi float [ [[TMP1]], %[[MIDDLE_BLOCK]] ], [ [[INIT]], %[[ENTRY]] ]
; INTERLEAVE-NEXT: br label %[[LOOP:.*]]
; INTERLEAVE: [[LOOP]]:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[FP_IV:%.*]] = phi float [ [[BC_RESUME_VAL2]], %[[SCALAR_PH]] ], [ [[ADD:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[GEP_A:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
; INTERLEAVE-NEXT: store float [[FP_IV]], ptr [[GEP_A]], align 4
; INTERLEAVE-NEXT: [[ADD]] = fadd fast float [[FP_IV]], [[FPINC]]
; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; INTERLEAVE-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; INTERLEAVE-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], {{!llvm.loop ![0-9]+}}
; INTERLEAVE: [[EXIT]]:
; INTERLEAVE-NEXT: [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], %[[LOOP]] ], [ [[TMP1]], %[[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: ret float [[ADD_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%fp.iv = phi float [ %init, %entry ], [ %add, %loop ]
%gep.A = getelementptr inbounds float, ptr %A, i64 %iv
store float %fp.iv, ptr %gep.A, align 4
%add = fadd fast float %fp.iv, %fpinc
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %N
br i1 %ec, label %exit, label %loop
exit:
ret float %add
}
define float @fp_postinc_use_fsub(float %init, ptr noalias nocapture %A, i64 %N, float %fpinc) {
; VEC-LABEL: define float @fp_postinc_use_fsub(
; VEC-SAME: float [[INIT:%.*]], ptr noalias nocapture [[A:%.*]], i64 [[N:%.*]], float [[FPINC:%.*]]) {
; VEC-NEXT: [[ENTRY:.*]]:
; VEC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 2
; VEC-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; VEC: [[VECTOR_PH]]:
; VEC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
; VEC-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; VEC-NEXT: [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
; VEC-NEXT: [[TMP0:%.*]] = fmul fast float [[FPINC]], [[DOTCAST]]
; VEC-NEXT: [[TMP1:%.*]] = fsub fast float [[INIT]], [[TMP0]]
; VEC-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[INIT]], i64 0
; VEC-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
; VEC-NEXT: [[DOTSPLATINSERT1:%.*]] = insertelement <2 x float> poison, float [[FPINC]], i64 0
; VEC-NEXT: [[DOTSPLAT2:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT1]], <2 x float> poison, <2 x i32> zeroinitializer
; VEC-NEXT: [[TMP2:%.*]] = fmul fast <2 x float> <float 0.000000e+00, float 1.000000e+00>, [[DOTSPLAT2]]
; VEC-NEXT: [[INDUCTION:%.*]] = fsub fast <2 x float> [[DOTSPLAT]], [[TMP2]]
; VEC-NEXT: [[TMP3:%.*]] = fmul fast float [[FPINC]], 2.000000e+00
; VEC-NEXT: [[DOTSPLATINSERT3:%.*]] = insertelement <2 x float> poison, float [[TMP3]], i64 0
; VEC-NEXT: [[DOTSPLAT4:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT3]], <2 x float> poison, <2 x i32> zeroinitializer
; VEC-NEXT: br label %[[VECTOR_BODY:.*]]
; VEC: [[VECTOR_BODY]]:
; VEC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VEC-NEXT: [[VEC_IND:%.*]] = phi <2 x float> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VEC-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 0
; VEC-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
; VEC-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
; VEC-NEXT: store <2 x float> [[VEC_IND]], ptr [[TMP6]], align 4
; VEC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; VEC-NEXT: [[VEC_IND_NEXT]] = fsub fast <2 x float> [[VEC_IND]], [[DOTSPLAT4]]
; VEC-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; VEC-NEXT: br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; VEC: [[MIDDLE_BLOCK]]:
; VEC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; VEC-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; VEC: [[SCALAR_PH]]:
; VEC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; VEC-NEXT: [[BC_RESUME_VAL5:%.*]] = phi float [ [[TMP1]], %[[MIDDLE_BLOCK]] ], [ [[INIT]], %[[ENTRY]] ]
; VEC-NEXT: br label %[[LOOP:.*]]
; VEC: [[LOOP]]:
; VEC-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; VEC-NEXT: [[FP_IV:%.*]] = phi float [ [[BC_RESUME_VAL5]], %[[SCALAR_PH]] ], [ [[ADD:%.*]], %[[LOOP]] ]
; VEC-NEXT: [[GEP_A:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
; VEC-NEXT: store float [[FP_IV]], ptr [[GEP_A]], align 4
; VEC-NEXT: [[ADD]] = fsub fast float [[FP_IV]], [[FPINC]]
; VEC-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; VEC-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; VEC-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], {{!llvm.loop ![0-9]+}}
; VEC: [[EXIT]]:
; VEC-NEXT: [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], %[[LOOP]] ], [ [[TMP1]], %[[MIDDLE_BLOCK]] ]
; VEC-NEXT: ret float [[ADD_LCSSA]]
;
; INTERLEAVE-LABEL: define float @fp_postinc_use_fsub(
; INTERLEAVE-SAME: float [[INIT:%.*]], ptr noalias nocapture [[A:%.*]], i64 [[N:%.*]], float [[FPINC:%.*]]) {
; INTERLEAVE-NEXT: [[ENTRY:.*]]:
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 2
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; INTERLEAVE: [[VECTOR_PH]]:
; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; INTERLEAVE-NEXT: [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
; INTERLEAVE-NEXT: [[TMP0:%.*]] = fmul fast float [[FPINC]], [[DOTCAST]]
; INTERLEAVE-NEXT: [[TMP1:%.*]] = fsub fast float [[INIT]], [[TMP0]]
; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]]
; INTERLEAVE: [[VECTOR_BODY]]:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 0
; INTERLEAVE-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 1
; INTERLEAVE-NEXT: [[DOTCAST1:%.*]] = sitofp i64 [[INDEX]] to float
; INTERLEAVE-NEXT: [[TMP4:%.*]] = fmul fast float [[FPINC]], [[DOTCAST1]]
; INTERLEAVE-NEXT: [[OFFSET_IDX:%.*]] = fsub fast float [[INIT]], [[TMP4]]
; INTERLEAVE-NEXT: [[TMP5:%.*]] = fmul fast float 0.000000e+00, [[FPINC]]
; INTERLEAVE-NEXT: [[TMP6:%.*]] = fsub fast float [[OFFSET_IDX]], [[TMP5]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = fmul fast float 1.000000e+00, [[FPINC]]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = fsub fast float [[OFFSET_IDX]], [[TMP7]]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP2]]
; INTERLEAVE-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP3]]
; INTERLEAVE-NEXT: store float [[TMP6]], ptr [[TMP9]], align 4
; INTERLEAVE-NEXT: store float [[TMP8]], ptr [[TMP10]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; INTERLEAVE-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
; INTERLEAVE: [[MIDDLE_BLOCK]]:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; INTERLEAVE: [[SCALAR_PH]]:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL2:%.*]] = phi float [ [[TMP1]], %[[MIDDLE_BLOCK]] ], [ [[INIT]], %[[ENTRY]] ]
; INTERLEAVE-NEXT: br label %[[LOOP:.*]]
; INTERLEAVE: [[LOOP]]:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[FP_IV:%.*]] = phi float [ [[BC_RESUME_VAL2]], %[[SCALAR_PH]] ], [ [[ADD:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[GEP_A:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
; INTERLEAVE-NEXT: store float [[FP_IV]], ptr [[GEP_A]], align 4
; INTERLEAVE-NEXT: [[ADD]] = fsub fast float [[FP_IV]], [[FPINC]]
; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; INTERLEAVE-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; INTERLEAVE-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], {{!llvm.loop ![0-9]+}}
; INTERLEAVE: [[EXIT]]:
; INTERLEAVE-NEXT: [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], %[[LOOP]] ], [ [[TMP1]], %[[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: ret float [[ADD_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%fp.iv = phi float [ %init, %entry ], [ %add, %loop ]
%gep.A = getelementptr inbounds float, ptr %A, i64 %iv
store float %fp.iv, ptr %gep.A, align 4
%add = fsub fast float %fp.iv, %fpinc
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %N
br i1 %ec, label %exit, label %loop
exit:
ret float %add
}
Reference in New Issue View Git Blame Copy Permalink