shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity
llvm-project/llvm/test/Transforms/LoopVectorize/PowerPC/optimal-epilog-vectorization.ll
Florian Hahn 77fb848894
Reapply "[LV] Simplify and unify resume value handling for epilogue vec." (#187504) 
This reverts commit cdaf29f84dd0abbd1f961982799059c92d76625b.

This version skips removeBranchOnConst when vectorizing the epilogue, as
it may trigger folds that remove the resume phi used as resume value
from the epilogue.

This fixes https://github.com/llvm/llvm-project/issues/187323.

Original message:
This patch tries to drastically simplify resume value handling for the
scalar loop when vectorizing the epilogue.

It uses a simpler, uniform approach for updating all resume values in
the scalar loop:

1. Create ResumeForEpilogue recipes for all scalar resume phis in the
main loop (the epilogue plan will have exactly the same scalar resume
phis, in exactly the same order)
2. Update ::execute for ResumeForEpilogue to set the underlying value
when executing. This is not super clean, but allows easy lookup of the
generated IR value when we update the resume phis in the epilogue. Once
we connect the 2 plans together explicitly, this can be removed.
3. Use the list of ResumeForEpilogue VPInstructions from the main loop
to update the resume/bypass values from the epilogue.

This simplifies the code quite a bit, makes it more robust (should fix
https://github.com/llvm/llvm-project/issues/179407) and also fixes a
mis-compile in the existing tests (see change in

llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-sub-epilogue-vec.ll,
where previously we would incorrectly resume using the start value when
the epilogue iteration check failed)

In some cases, we get simpler code, due to additional CSE, in some cases
the induction end value computations get moved from the epilogue
iteration check to the vector preheader. We could try to sink the
instructions as cleanup, but it is probably not worth the trouble.

Fixes https://github.com/llvm/llvm-project/issues/179407.

PR for recommit https://github.com/llvm/llvm-project/pull/188134
2026-03-23 22:09:40 +00:00

486 lines
38 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "scalar.ph:" --version 5
; RUN: opt < %s -passes='loop-vectorize' -enable-epilogue-vectorization -epilogue-vectorization-force-VF=2 -S | FileCheck %s --check-prefix VF-TWO-CHECK
; RUN: opt < %s -passes='loop-vectorize' -enable-epilogue-vectorization -epilogue-vectorization-force-VF=4 -S | FileCheck %s --check-prefix VF-FOUR-CHECK
target datalayout = "e-m:e-i64:64-n32:64"
target triple = "powerpc64le-unknown-linux-gnu"
define void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32 %N) {
; VF-TWO-CHECK-LABEL: define void @f1(
; VF-TWO-CHECK-SAME: ptr noalias [[AA:%.*]], ptr noalias [[BB:%.*]], ptr noalias [[CC:%.*]], i32 [[N:%.*]]) {
; VF-TWO-CHECK-NEXT: [[ENTRY:.*:]]
; VF-TWO-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64
; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 2
; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[VEC_EPILOG_SCALAR_PH:.*]], label %[[VECTOR_MAIN_LOOP_ITER_CHECK:.*]]
; VF-TWO-CHECK: [[VECTOR_MAIN_LOOP_ITER_CHECK]]:
; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 32
; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label %[[VEC_EPILOG_PH:.*]], label %[[VECTOR_PH:.*]]
; VF-TWO-CHECK: [[VECTOR_PH]]:
; VF-TWO-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 32
; VF-TWO-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]]
; VF-TWO-CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; VF-TWO-CHECK: [[VECTOR_BODY]]:
; VF-TWO-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VF-TWO-CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[BB]], i64 [[INDEX]]
; VF-TWO-CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 4
; VF-TWO-CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 8
; VF-TWO-CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 12
; VF-TWO-CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 16
; VF-TWO-CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 20
; VF-TWO-CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 24
; VF-TWO-CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 28
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP8]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP17]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <4 x float>, ptr [[TMP18]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP19]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP20]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP21]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP22]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP23]], align 4
; VF-TWO-CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds float, ptr [[CC]], i64 [[INDEX]]
; VF-TWO-CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 4
; VF-TWO-CHECK-NEXT: [[TMP34:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 8
; VF-TWO-CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 12
; VF-TWO-CHECK-NEXT: [[TMP36:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 16
; VF-TWO-CHECK-NEXT: [[TMP37:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 20
; VF-TWO-CHECK-NEXT: [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 24
; VF-TWO-CHECK-NEXT: [[TMP39:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 28
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP24]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, ptr [[TMP33]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD11:%.*]] = load <4 x float>, ptr [[TMP34]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, ptr [[TMP35]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD13:%.*]] = load <4 x float>, ptr [[TMP36]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, ptr [[TMP37]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD15:%.*]] = load <4 x float>, ptr [[TMP38]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD16:%.*]] = load <4 x float>, ptr [[TMP39]], align 4
; VF-TWO-CHECK-NEXT: [[TMP40:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD9]]
; VF-TWO-CHECK-NEXT: [[TMP41:%.*]] = fadd fast <4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD10]]
; VF-TWO-CHECK-NEXT: [[TMP42:%.*]] = fadd fast <4 x float> [[WIDE_LOAD3]], [[WIDE_LOAD11]]
; VF-TWO-CHECK-NEXT: [[TMP43:%.*]] = fadd fast <4 x float> [[WIDE_LOAD4]], [[WIDE_LOAD12]]
; VF-TWO-CHECK-NEXT: [[TMP44:%.*]] = fadd fast <4 x float> [[WIDE_LOAD5]], [[WIDE_LOAD13]]
; VF-TWO-CHECK-NEXT: [[TMP45:%.*]] = fadd fast <4 x float> [[WIDE_LOAD6]], [[WIDE_LOAD14]]
; VF-TWO-CHECK-NEXT: [[TMP46:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD15]]
; VF-TWO-CHECK-NEXT: [[TMP47:%.*]] = fadd fast <4 x float> [[WIDE_LOAD8]], [[WIDE_LOAD16]]
; VF-TWO-CHECK-NEXT: [[TMP48:%.*]] = getelementptr inbounds float, ptr [[AA]], i64 [[INDEX]]
; VF-TWO-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 4
; VF-TWO-CHECK-NEXT: [[TMP58:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 8
; VF-TWO-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 12
; VF-TWO-CHECK-NEXT: [[TMP60:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 16
; VF-TWO-CHECK-NEXT: [[TMP61:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 20
; VF-TWO-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 24
; VF-TWO-CHECK-NEXT: [[TMP63:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 28
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP40]], ptr [[TMP48]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP41]], ptr [[TMP57]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP42]], ptr [[TMP58]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP43]], ptr [[TMP59]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP44]], ptr [[TMP60]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP45]], ptr [[TMP61]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP46]], ptr [[TMP62]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP47]], ptr [[TMP63]], align 4
; VF-TWO-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
; VF-TWO-CHECK-NEXT: [[TMP64:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; VF-TWO-CHECK-NEXT: br i1 [[TMP64]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; VF-TWO-CHECK: [[MIDDLE_BLOCK]]:
; VF-TWO-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; VF-TWO-CHECK-NEXT: br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[VEC_EPILOG_ITER_CHECK:.*]]
; VF-TWO-CHECK: [[VEC_EPILOG_ITER_CHECK]]:
; VF-TWO-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_MOD_VF]], 2
; VF-TWO-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label %[[VEC_EPILOG_SCALAR_PH]], label %[[VEC_EPILOG_PH]], !prof [[PROF3:![0-9]+]]
; VF-TWO-CHECK: [[VEC_EPILOG_PH]]:
; VF-TWO-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; VF-TWO-CHECK-NEXT: [[N_MOD_VF17:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 2
; VF-TWO-CHECK-NEXT: [[N_VEC18:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF17]]
; VF-TWO-CHECK-NEXT: br label %[[VEC_EPILOG_VECTOR_BODY:.*]]
; VF-TWO-CHECK: [[VEC_EPILOG_VECTOR_BODY]]:
; VF-TWO-CHECK-NEXT: [[INDEX20:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT23:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
; VF-TWO-CHECK-NEXT: [[TMP66:%.*]] = getelementptr inbounds float, ptr [[BB]], i64 [[INDEX20]]
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD21:%.*]] = load <2 x float>, ptr [[TMP66]], align 4
; VF-TWO-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, ptr [[CC]], i64 [[INDEX20]]
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD22:%.*]] = load <2 x float>, ptr [[TMP68]], align 4
; VF-TWO-CHECK-NEXT: [[TMP70:%.*]] = fadd fast <2 x float> [[WIDE_LOAD21]], [[WIDE_LOAD22]]
; VF-TWO-CHECK-NEXT: [[TMP71:%.*]] = getelementptr inbounds float, ptr [[AA]], i64 [[INDEX20]]
; VF-TWO-CHECK-NEXT: store <2 x float> [[TMP70]], ptr [[TMP71]], align 4
; VF-TWO-CHECK-NEXT: [[INDEX_NEXT23]] = add nuw i64 [[INDEX20]], 2
; VF-TWO-CHECK-NEXT: [[TMP73:%.*]] = icmp eq i64 [[INDEX_NEXT23]], [[N_VEC18]]
; VF-TWO-CHECK-NEXT: br i1 [[TMP73]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; VF-TWO-CHECK: [[VEC_EPILOG_MIDDLE_BLOCK]]:
; VF-TWO-CHECK-NEXT: [[CMP_N19:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC18]]
; VF-TWO-CHECK-NEXT: br i1 [[CMP_N19]], [[EXIT]], label %[[VEC_EPILOG_SCALAR_PH]]
; VF-TWO-CHECK: [[VEC_EPILOG_SCALAR_PH]]:
;
; VF-FOUR-CHECK-LABEL: define void @f1(
; VF-FOUR-CHECK-SAME: ptr noalias [[AA:%.*]], ptr noalias [[BB:%.*]], ptr noalias [[CC:%.*]], i32 [[N:%.*]]) {
; VF-FOUR-CHECK-NEXT: [[ENTRY:.*:]]
; VF-FOUR-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64
; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4
; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[VEC_EPILOG_SCALAR_PH:.*]], label %[[VECTOR_MAIN_LOOP_ITER_CHECK:.*]]
; VF-FOUR-CHECK: [[VECTOR_MAIN_LOOP_ITER_CHECK]]:
; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 32
; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label %[[VEC_EPILOG_PH:.*]], label %[[VECTOR_PH:.*]]
; VF-FOUR-CHECK: [[VECTOR_PH]]:
; VF-FOUR-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 32
; VF-FOUR-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]]
; VF-FOUR-CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; VF-FOUR-CHECK: [[VECTOR_BODY]]:
; VF-FOUR-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VF-FOUR-CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[BB]], i64 [[INDEX]]
; VF-FOUR-CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 4
; VF-FOUR-CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 8
; VF-FOUR-CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 12
; VF-FOUR-CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 16
; VF-FOUR-CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 20
; VF-FOUR-CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 24
; VF-FOUR-CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 28
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP8]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP17]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <4 x float>, ptr [[TMP18]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP19]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP20]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP21]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP22]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP23]], align 4
; VF-FOUR-CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds float, ptr [[CC]], i64 [[INDEX]]
; VF-FOUR-CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 4
; VF-FOUR-CHECK-NEXT: [[TMP34:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 8
; VF-FOUR-CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 12
; VF-FOUR-CHECK-NEXT: [[TMP36:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 16
; VF-FOUR-CHECK-NEXT: [[TMP37:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 20
; VF-FOUR-CHECK-NEXT: [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 24
; VF-FOUR-CHECK-NEXT: [[TMP39:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i64 28
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP24]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, ptr [[TMP33]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD11:%.*]] = load <4 x float>, ptr [[TMP34]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, ptr [[TMP35]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD13:%.*]] = load <4 x float>, ptr [[TMP36]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, ptr [[TMP37]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD15:%.*]] = load <4 x float>, ptr [[TMP38]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD16:%.*]] = load <4 x float>, ptr [[TMP39]], align 4
; VF-FOUR-CHECK-NEXT: [[TMP40:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD9]]
; VF-FOUR-CHECK-NEXT: [[TMP41:%.*]] = fadd fast <4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD10]]
; VF-FOUR-CHECK-NEXT: [[TMP42:%.*]] = fadd fast <4 x float> [[WIDE_LOAD3]], [[WIDE_LOAD11]]
; VF-FOUR-CHECK-NEXT: [[TMP43:%.*]] = fadd fast <4 x float> [[WIDE_LOAD4]], [[WIDE_LOAD12]]
; VF-FOUR-CHECK-NEXT: [[TMP44:%.*]] = fadd fast <4 x float> [[WIDE_LOAD5]], [[WIDE_LOAD13]]
; VF-FOUR-CHECK-NEXT: [[TMP45:%.*]] = fadd fast <4 x float> [[WIDE_LOAD6]], [[WIDE_LOAD14]]
; VF-FOUR-CHECK-NEXT: [[TMP46:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD15]]
; VF-FOUR-CHECK-NEXT: [[TMP47:%.*]] = fadd fast <4 x float> [[WIDE_LOAD8]], [[WIDE_LOAD16]]
; VF-FOUR-CHECK-NEXT: [[TMP48:%.*]] = getelementptr inbounds float, ptr [[AA]], i64 [[INDEX]]
; VF-FOUR-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 4
; VF-FOUR-CHECK-NEXT: [[TMP58:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 8
; VF-FOUR-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 12
; VF-FOUR-CHECK-NEXT: [[TMP60:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 16
; VF-FOUR-CHECK-NEXT: [[TMP61:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 20
; VF-FOUR-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 24
; VF-FOUR-CHECK-NEXT: [[TMP63:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 28
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP40]], ptr [[TMP48]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP41]], ptr [[TMP57]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP42]], ptr [[TMP58]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP43]], ptr [[TMP59]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP44]], ptr [[TMP60]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP45]], ptr [[TMP61]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP46]], ptr [[TMP62]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP47]], ptr [[TMP63]], align 4
; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
; VF-FOUR-CHECK-NEXT: [[TMP64:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; VF-FOUR-CHECK-NEXT: br i1 [[TMP64]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; VF-FOUR-CHECK: [[MIDDLE_BLOCK]]:
; VF-FOUR-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[VEC_EPILOG_ITER_CHECK:.*]]
; VF-FOUR-CHECK: [[VEC_EPILOG_ITER_CHECK]]:
; VF-FOUR-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_MOD_VF]], 4
; VF-FOUR-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label %[[VEC_EPILOG_SCALAR_PH]], label %[[VEC_EPILOG_PH]], !prof [[PROF3:![0-9]+]]
; VF-FOUR-CHECK: [[VEC_EPILOG_PH]]:
; VF-FOUR-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; VF-FOUR-CHECK-NEXT: [[N_MOD_VF17:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
; VF-FOUR-CHECK-NEXT: [[N_VEC18:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF17]]
; VF-FOUR-CHECK-NEXT: br label %[[VEC_EPILOG_VECTOR_BODY:.*]]
; VF-FOUR-CHECK: [[VEC_EPILOG_VECTOR_BODY]]:
; VF-FOUR-CHECK-NEXT: [[INDEX20:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT23:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
; VF-FOUR-CHECK-NEXT: [[TMP66:%.*]] = getelementptr inbounds float, ptr [[BB]], i64 [[INDEX20]]
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD21:%.*]] = load <4 x float>, ptr [[TMP66]], align 4
; VF-FOUR-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, ptr [[CC]], i64 [[INDEX20]]
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD22:%.*]] = load <4 x float>, ptr [[TMP68]], align 4
; VF-FOUR-CHECK-NEXT: [[TMP70:%.*]] = fadd fast <4 x float> [[WIDE_LOAD21]], [[WIDE_LOAD22]]
; VF-FOUR-CHECK-NEXT: [[TMP71:%.*]] = getelementptr inbounds float, ptr [[AA]], i64 [[INDEX20]]
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP70]], ptr [[TMP71]], align 4
; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT23]] = add nuw i64 [[INDEX20]], 4
; VF-FOUR-CHECK-NEXT: [[TMP73:%.*]] = icmp eq i64 [[INDEX_NEXT23]], [[N_VEC18]]
; VF-FOUR-CHECK-NEXT: br i1 [[TMP73]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; VF-FOUR-CHECK: [[VEC_EPILOG_MIDDLE_BLOCK]]:
; VF-FOUR-CHECK-NEXT: [[CMP_N19:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC18]]
; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N19]], [[EXIT]], label %[[VEC_EPILOG_SCALAR_PH]]
; VF-FOUR-CHECK: [[VEC_EPILOG_SCALAR_PH]]:
;
entry:
%n.ext = zext i32 %N to i64
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%arrayidx = getelementptr inbounds float, ptr %bb, i64 %iv
%0 = load float, ptr %arrayidx, align 4
%arrayidx2 = getelementptr inbounds float, ptr %cc, i64 %iv
%1 = load float, ptr %arrayidx2, align 4
%add = fadd fast float %0, %1
%arrayidx4 = getelementptr inbounds float, ptr %aa, i64 %iv
store float %add, ptr %arrayidx4, align 4
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp ne i64 %iv.next, %n.ext
br i1 %ec, label %loop, label %exit
exit:
ret void
}
define void @f2(ptr noalias %A, ptr noalias %B, i32 %n) {
; VF-TWO-CHECK-LABEL: define void @f2(
; VF-TWO-CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i32 [[N:%.*]]) {
; VF-TWO-CHECK-NEXT: [[ITER_CHECK:.*:]]
; VF-TWO-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64
; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 2
; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[VEC_EPILOG_SCALAR_PH:.*]], label %[[VECTOR_SCEVCHECK:.*]]
; VF-TWO-CHECK: [[VECTOR_SCEVCHECK]]:
; VF-TWO-CHECK-NEXT: [[TMP1:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1
; VF-TWO-CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1
; VF-TWO-CHECK-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
; VF-TWO-CHECK-NEXT: [[TMP3:%.*]] = sub i32 [[TMP0]], [[TMP2]]
; VF-TWO-CHECK-NEXT: [[TMP4:%.*]] = icmp sgt i32 [[TMP3]], [[TMP0]]
; VF-TWO-CHECK-NEXT: [[TMP6:%.*]] = icmp ugt i64 [[TMP1]], 4294967295
; VF-TWO-CHECK-NEXT: [[TMP7:%.*]] = or i1 [[TMP4]], [[TMP6]]
; VF-TWO-CHECK-NEXT: br i1 [[TMP7]], label %[[VEC_EPILOG_SCALAR_PH]], label %[[VECTOR_MAIN_LOOP_ITER_CHECK:.*]]
; VF-TWO-CHECK: [[VECTOR_MAIN_LOOP_ITER_CHECK]]:
; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 32
; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label %[[VEC_EPILOG_PH:.*]], label %[[VECTOR_PH:.*]]
; VF-TWO-CHECK: [[VECTOR_PH]]:
; VF-TWO-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 32
; VF-TWO-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]]
; VF-TWO-CHECK-NEXT: [[IND_END18:%.*]] = trunc i64 [[N_VEC]] to i32
; VF-TWO-CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; VF-TWO-CHECK: [[VECTOR_BODY]]:
; VF-TWO-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VF-TWO-CHECK-NEXT: [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
; VF-TWO-CHECK-NEXT: [[TMP24:%.*]] = xor i32 [[OFFSET_IDX]], -1
; VF-TWO-CHECK-NEXT: [[TMP32:%.*]] = add i32 [[TMP24]], [[N]]
; VF-TWO-CHECK-NEXT: [[TMP40:%.*]] = sext i32 [[TMP32]] to i64
; VF-TWO-CHECK-NEXT: [[TMP48:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP40]]
; VF-TWO-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -3
; VF-TWO-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -7
; VF-TWO-CHECK-NEXT: [[TMP61:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -11
; VF-TWO-CHECK-NEXT: [[TMP63:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -15
; VF-TWO-CHECK-NEXT: [[TMP65:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -19
; VF-TWO-CHECK-NEXT: [[TMP67:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -23
; VF-TWO-CHECK-NEXT: [[TMP69:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -27
; VF-TWO-CHECK-NEXT: [[TMP71:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -31
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP57]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP59]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP61]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP63]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP65]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, ptr [[TMP67]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, ptr [[TMP69]], align 4
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, ptr [[TMP71]], align 4
; VF-TWO-CHECK-NEXT: [[REVERSE:%.*]] = shufflevector <4 x float> [[WIDE_LOAD]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-TWO-CHECK-NEXT: [[REVERSE3:%.*]] = shufflevector <4 x float> [[WIDE_LOAD2]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-TWO-CHECK-NEXT: [[REVERSE5:%.*]] = shufflevector <4 x float> [[WIDE_LOAD4]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-TWO-CHECK-NEXT: [[REVERSE7:%.*]] = shufflevector <4 x float> [[WIDE_LOAD6]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-TWO-CHECK-NEXT: [[REVERSE9:%.*]] = shufflevector <4 x float> [[WIDE_LOAD8]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-TWO-CHECK-NEXT: [[REVERSE11:%.*]] = shufflevector <4 x float> [[WIDE_LOAD10]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-TWO-CHECK-NEXT: [[REVERSE13:%.*]] = shufflevector <4 x float> [[WIDE_LOAD12]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-TWO-CHECK-NEXT: [[REVERSE15:%.*]] = shufflevector <4 x float> [[WIDE_LOAD14]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-TWO-CHECK-NEXT: [[TMP72:%.*]] = fadd fast <4 x float> [[REVERSE]], splat (float 1.000000e+00)
; VF-TWO-CHECK-NEXT: [[TMP73:%.*]] = fadd fast <4 x float> [[REVERSE3]], splat (float 1.000000e+00)
; VF-TWO-CHECK-NEXT: [[TMP74:%.*]] = fadd fast <4 x float> [[REVERSE5]], splat (float 1.000000e+00)
; VF-TWO-CHECK-NEXT: [[TMP75:%.*]] = fadd fast <4 x float> [[REVERSE7]], splat (float 1.000000e+00)
; VF-TWO-CHECK-NEXT: [[TMP76:%.*]] = fadd fast <4 x float> [[REVERSE9]], splat (float 1.000000e+00)
; VF-TWO-CHECK-NEXT: [[TMP77:%.*]] = fadd fast <4 x float> [[REVERSE11]], splat (float 1.000000e+00)
; VF-TWO-CHECK-NEXT: [[TMP78:%.*]] = fadd fast <4 x float> [[REVERSE13]], splat (float 1.000000e+00)
; VF-TWO-CHECK-NEXT: [[TMP79:%.*]] = fadd fast <4 x float> [[REVERSE15]], splat (float 1.000000e+00)
; VF-TWO-CHECK-NEXT: [[TMP80:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
; VF-TWO-CHECK-NEXT: [[TMP89:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 4
; VF-TWO-CHECK-NEXT: [[TMP90:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 8
; VF-TWO-CHECK-NEXT: [[TMP91:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 12
; VF-TWO-CHECK-NEXT: [[TMP92:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 16
; VF-TWO-CHECK-NEXT: [[TMP93:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 20
; VF-TWO-CHECK-NEXT: [[TMP94:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 24
; VF-TWO-CHECK-NEXT: [[TMP95:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 28
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP72]], ptr [[TMP80]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP73]], ptr [[TMP89]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP74]], ptr [[TMP90]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP75]], ptr [[TMP91]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP76]], ptr [[TMP92]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP77]], ptr [[TMP93]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP78]], ptr [[TMP94]], align 4
; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP79]], ptr [[TMP95]], align 4
; VF-TWO-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
; VF-TWO-CHECK-NEXT: [[TMP96:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; VF-TWO-CHECK-NEXT: br i1 [[TMP96]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; VF-TWO-CHECK: [[MIDDLE_BLOCK]]:
; VF-TWO-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; VF-TWO-CHECK-NEXT: br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[VEC_EPILOG_ITER_CHECK:.*]]
; VF-TWO-CHECK: [[VEC_EPILOG_ITER_CHECK]]:
; VF-TWO-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_MOD_VF]], 2
; VF-TWO-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label %[[VEC_EPILOG_SCALAR_PH]], label %[[VEC_EPILOG_PH]], !prof [[PROF3]]
; VF-TWO-CHECK: [[VEC_EPILOG_PH]]:
; VF-TWO-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; VF-TWO-CHECK-NEXT: [[N_MOD_VF16:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 2
; VF-TWO-CHECK-NEXT: [[N_VEC17:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF16]]
; VF-TWO-CHECK-NEXT: [[IND_END:%.*]] = trunc i64 [[N_VEC17]] to i32
; VF-TWO-CHECK-NEXT: br label %[[VEC_EPILOG_VECTOR_BODY:.*]]
; VF-TWO-CHECK: [[VEC_EPILOG_VECTOR_BODY]]:
; VF-TWO-CHECK-NEXT: [[INDEX21:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT25:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
; VF-TWO-CHECK-NEXT: [[OFFSET_IDX22:%.*]] = trunc i64 [[INDEX21]] to i32
; VF-TWO-CHECK-NEXT: [[TMP99:%.*]] = xor i32 [[OFFSET_IDX22]], -1
; VF-TWO-CHECK-NEXT: [[TMP100:%.*]] = add i32 [[TMP99]], [[N]]
; VF-TWO-CHECK-NEXT: [[TMP101:%.*]] = sext i32 [[TMP100]] to i64
; VF-TWO-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP101]]
; VF-TWO-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, ptr [[TMP102]], i64 -1
; VF-TWO-CHECK-NEXT: [[WIDE_LOAD23:%.*]] = load <2 x float>, ptr [[TMP104]], align 4
; VF-TWO-CHECK-NEXT: [[REVERSE24:%.*]] = shufflevector <2 x float> [[WIDE_LOAD23]], <2 x float> poison, <2 x i32> <i32 1, i32 0>
; VF-TWO-CHECK-NEXT: [[TMP105:%.*]] = fadd fast <2 x float> [[REVERSE24]], splat (float 1.000000e+00)
; VF-TWO-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX21]]
; VF-TWO-CHECK-NEXT: store <2 x float> [[TMP105]], ptr [[TMP106]], align 4
; VF-TWO-CHECK-NEXT: [[INDEX_NEXT25]] = add nuw i64 [[INDEX21]], 2
; VF-TWO-CHECK-NEXT: [[TMP108:%.*]] = icmp eq i64 [[INDEX_NEXT25]], [[N_VEC17]]
; VF-TWO-CHECK-NEXT: br i1 [[TMP108]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
; VF-TWO-CHECK: [[VEC_EPILOG_MIDDLE_BLOCK]]:
; VF-TWO-CHECK-NEXT: [[CMP_N20:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC17]]
; VF-TWO-CHECK-NEXT: br i1 [[CMP_N20]], [[EXIT]], label %[[VEC_EPILOG_SCALAR_PH]]
; VF-TWO-CHECK: [[VEC_EPILOG_SCALAR_PH]]:
;
; VF-FOUR-CHECK-LABEL: define void @f2(
; VF-FOUR-CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i32 [[N:%.*]]) {
; VF-FOUR-CHECK-NEXT: [[ITER_CHECK:.*:]]
; VF-FOUR-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64
; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4
; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[VEC_EPILOG_SCALAR_PH:.*]], label %[[VECTOR_SCEVCHECK:.*]]
; VF-FOUR-CHECK: [[VECTOR_SCEVCHECK]]:
; VF-FOUR-CHECK-NEXT: [[TMP1:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1
; VF-FOUR-CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1
; VF-FOUR-CHECK-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
; VF-FOUR-CHECK-NEXT: [[TMP3:%.*]] = sub i32 [[TMP0]], [[TMP2]]
; VF-FOUR-CHECK-NEXT: [[TMP4:%.*]] = icmp sgt i32 [[TMP3]], [[TMP0]]
; VF-FOUR-CHECK-NEXT: [[TMP6:%.*]] = icmp ugt i64 [[TMP1]], 4294967295
; VF-FOUR-CHECK-NEXT: [[TMP7:%.*]] = or i1 [[TMP4]], [[TMP6]]
; VF-FOUR-CHECK-NEXT: br i1 [[TMP7]], label %[[VEC_EPILOG_SCALAR_PH]], label %[[VECTOR_MAIN_LOOP_ITER_CHECK:.*]]
; VF-FOUR-CHECK: [[VECTOR_MAIN_LOOP_ITER_CHECK]]:
; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 32
; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label %[[VEC_EPILOG_PH:.*]], label %[[VECTOR_PH:.*]]
; VF-FOUR-CHECK: [[VECTOR_PH]]:
; VF-FOUR-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 32
; VF-FOUR-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]]
; VF-FOUR-CHECK-NEXT: [[IND_END18:%.*]] = trunc i64 [[N_VEC]] to i32
; VF-FOUR-CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; VF-FOUR-CHECK: [[VECTOR_BODY]]:
; VF-FOUR-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; VF-FOUR-CHECK-NEXT: [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
; VF-FOUR-CHECK-NEXT: [[TMP24:%.*]] = xor i32 [[OFFSET_IDX]], -1
; VF-FOUR-CHECK-NEXT: [[TMP32:%.*]] = add i32 [[TMP24]], [[N]]
; VF-FOUR-CHECK-NEXT: [[TMP40:%.*]] = sext i32 [[TMP32]] to i64
; VF-FOUR-CHECK-NEXT: [[TMP48:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP40]]
; VF-FOUR-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -3
; VF-FOUR-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -7
; VF-FOUR-CHECK-NEXT: [[TMP61:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -11
; VF-FOUR-CHECK-NEXT: [[TMP63:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -15
; VF-FOUR-CHECK-NEXT: [[TMP65:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -19
; VF-FOUR-CHECK-NEXT: [[TMP67:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -23
; VF-FOUR-CHECK-NEXT: [[TMP69:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -27
; VF-FOUR-CHECK-NEXT: [[TMP71:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 -31
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP57]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP59]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP61]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP63]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP65]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, ptr [[TMP67]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, ptr [[TMP69]], align 4
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, ptr [[TMP71]], align 4
; VF-FOUR-CHECK-NEXT: [[REVERSE:%.*]] = shufflevector <4 x float> [[WIDE_LOAD]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-FOUR-CHECK-NEXT: [[REVERSE3:%.*]] = shufflevector <4 x float> [[WIDE_LOAD2]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-FOUR-CHECK-NEXT: [[REVERSE5:%.*]] = shufflevector <4 x float> [[WIDE_LOAD4]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-FOUR-CHECK-NEXT: [[REVERSE7:%.*]] = shufflevector <4 x float> [[WIDE_LOAD6]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-FOUR-CHECK-NEXT: [[REVERSE9:%.*]] = shufflevector <4 x float> [[WIDE_LOAD8]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-FOUR-CHECK-NEXT: [[REVERSE11:%.*]] = shufflevector <4 x float> [[WIDE_LOAD10]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-FOUR-CHECK-NEXT: [[REVERSE13:%.*]] = shufflevector <4 x float> [[WIDE_LOAD12]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-FOUR-CHECK-NEXT: [[REVERSE15:%.*]] = shufflevector <4 x float> [[WIDE_LOAD14]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-FOUR-CHECK-NEXT: [[TMP72:%.*]] = fadd fast <4 x float> [[REVERSE]], splat (float 1.000000e+00)
; VF-FOUR-CHECK-NEXT: [[TMP73:%.*]] = fadd fast <4 x float> [[REVERSE3]], splat (float 1.000000e+00)
; VF-FOUR-CHECK-NEXT: [[TMP74:%.*]] = fadd fast <4 x float> [[REVERSE5]], splat (float 1.000000e+00)
; VF-FOUR-CHECK-NEXT: [[TMP75:%.*]] = fadd fast <4 x float> [[REVERSE7]], splat (float 1.000000e+00)
; VF-FOUR-CHECK-NEXT: [[TMP76:%.*]] = fadd fast <4 x float> [[REVERSE9]], splat (float 1.000000e+00)
; VF-FOUR-CHECK-NEXT: [[TMP77:%.*]] = fadd fast <4 x float> [[REVERSE11]], splat (float 1.000000e+00)
; VF-FOUR-CHECK-NEXT: [[TMP78:%.*]] = fadd fast <4 x float> [[REVERSE13]], splat (float 1.000000e+00)
; VF-FOUR-CHECK-NEXT: [[TMP79:%.*]] = fadd fast <4 x float> [[REVERSE15]], splat (float 1.000000e+00)
; VF-FOUR-CHECK-NEXT: [[TMP80:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
; VF-FOUR-CHECK-NEXT: [[TMP89:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 4
; VF-FOUR-CHECK-NEXT: [[TMP90:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 8
; VF-FOUR-CHECK-NEXT: [[TMP91:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 12
; VF-FOUR-CHECK-NEXT: [[TMP92:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 16
; VF-FOUR-CHECK-NEXT: [[TMP93:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 20
; VF-FOUR-CHECK-NEXT: [[TMP94:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 24
; VF-FOUR-CHECK-NEXT: [[TMP95:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i64 28
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP72]], ptr [[TMP80]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP73]], ptr [[TMP89]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP74]], ptr [[TMP90]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP75]], ptr [[TMP91]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP76]], ptr [[TMP92]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP77]], ptr [[TMP93]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP78]], ptr [[TMP94]], align 4
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP79]], ptr [[TMP95]], align 4
; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
; VF-FOUR-CHECK-NEXT: [[TMP96:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; VF-FOUR-CHECK-NEXT: br i1 [[TMP96]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; VF-FOUR-CHECK: [[MIDDLE_BLOCK]]:
; VF-FOUR-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[VEC_EPILOG_ITER_CHECK:.*]]
; VF-FOUR-CHECK: [[VEC_EPILOG_ITER_CHECK]]:
; VF-FOUR-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_MOD_VF]], 4
; VF-FOUR-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label %[[VEC_EPILOG_SCALAR_PH]], label %[[VEC_EPILOG_PH]], !prof [[PROF3]]
; VF-FOUR-CHECK: [[VEC_EPILOG_PH]]:
; VF-FOUR-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; VF-FOUR-CHECK-NEXT: [[N_MOD_VF16:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
; VF-FOUR-CHECK-NEXT: [[N_VEC17:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF16]]
; VF-FOUR-CHECK-NEXT: [[IND_END:%.*]] = trunc i64 [[N_VEC17]] to i32
; VF-FOUR-CHECK-NEXT: br label %[[VEC_EPILOG_VECTOR_BODY:.*]]
; VF-FOUR-CHECK: [[VEC_EPILOG_VECTOR_BODY]]:
; VF-FOUR-CHECK-NEXT: [[INDEX21:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT25:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
; VF-FOUR-CHECK-NEXT: [[OFFSET_IDX22:%.*]] = trunc i64 [[INDEX21]] to i32
; VF-FOUR-CHECK-NEXT: [[TMP99:%.*]] = xor i32 [[OFFSET_IDX22]], -1
; VF-FOUR-CHECK-NEXT: [[TMP100:%.*]] = add i32 [[TMP99]], [[N]]
; VF-FOUR-CHECK-NEXT: [[TMP101:%.*]] = sext i32 [[TMP100]] to i64
; VF-FOUR-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP101]]
; VF-FOUR-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, ptr [[TMP102]], i64 -3
; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD23:%.*]] = load <4 x float>, ptr [[TMP104]], align 4
; VF-FOUR-CHECK-NEXT: [[REVERSE24:%.*]] = shufflevector <4 x float> [[WIDE_LOAD23]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; VF-FOUR-CHECK-NEXT: [[TMP105:%.*]] = fadd fast <4 x float> [[REVERSE24]], splat (float 1.000000e+00)
; VF-FOUR-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX21]]
; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP105]], ptr [[TMP106]], align 4
; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT25]] = add nuw i64 [[INDEX21]], 4
; VF-FOUR-CHECK-NEXT: [[TMP108:%.*]] = icmp eq i64 [[INDEX_NEXT25]], [[N_VEC17]]
; VF-FOUR-CHECK-NEXT: br i1 [[TMP108]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
; VF-FOUR-CHECK: [[VEC_EPILOG_MIDDLE_BLOCK]]:
; VF-FOUR-CHECK-NEXT: [[CMP_N20:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC17]]
; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N20]], [[EXIT]], label %[[VEC_EPILOG_SCALAR_PH]]
; VF-FOUR-CHECK: [[VEC_EPILOG_SCALAR_PH]]:
;
entry:
%n.ext = zext i32 %n to i64
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%iv.2 = phi i32 [ 0, %entry ], [ %iv.2.next, %loop ]
%1 = xor i32 %iv.2, -1
%sub2 = add i32 %1, %n
%idxprom = sext i32 %sub2 to i64
%arrayidx = getelementptr inbounds float, ptr %B, i64 %idxprom
%2 = load float, ptr %arrayidx, align 4
%conv3 = fadd fast float %2, 1.000000e+00
%gep.A = getelementptr inbounds float, ptr %A, i64 %iv
store float %conv3, ptr %gep.A, align 4
%iv.next = add nuw nsw i64 %iv, 1
%iv.2.next = add nuw nsw i32 %iv.2, 1
%ec = icmp ne i64 %iv.next, %n.ext
br i1 %ec, label %loop, label %exit
exit:
ret void
}
Reference in New Issue View Git Blame Copy Permalink