[llvm] [LoopVectorize] Avoid considering PHI without any usage. (PR #92399)
Dinar Temirbulatov via llvm-commits
llvm-commits at lists.llvm.org
Thu May 16 06:30:04 PDT 2024
https://github.com/dtemirbulatov created https://github.com/llvm/llvm-project/pull/92399
We consider PHIs without any usage in the loop. This might result in generating extra unnecessary code, for example, vector splices.
>From 78bc8eb6d223ef524c0727746e11013ef38f79cd Mon Sep 17 00:00:00 2001
From: Dinar Temirbulatov <Dinar.Temirbulatov at arm.com>
Date: Thu, 16 May 2024 13:21:21 +0000
Subject: [PATCH] [LoopVectorize] Avoid considering PHI without any usage. We
consider PHIs without any usage in the loop. This might result in generating
extra unnecessary code, for example, vector splices.
---
llvm/lib/Analysis/IVDescriptors.cpp | 4 +-
.../Vectorize/LoopVectorizationLegality.cpp | 3 +
.../Transforms/Vectorize/LoopVectorize.cpp | 3 +
.../AArch64/loop-vectorization-factors.ll | 78 ++++++++++++-------
4 files changed, 61 insertions(+), 27 deletions(-)
diff --git a/llvm/lib/Analysis/IVDescriptors.cpp b/llvm/lib/Analysis/IVDescriptors.cpp
index 055f121e74341..49193474640ce 100644
--- a/llvm/lib/Analysis/IVDescriptors.cpp
+++ b/llvm/lib/Analysis/IVDescriptors.cpp
@@ -642,7 +642,6 @@ RecurrenceDescriptor::isAnyOfPattern(Loop *Loop, PHINode *OrigPhi,
SelectInst *SI = cast<SelectInst>(I);
Value *NonPhi = nullptr;
-
if (OrigPhi == dyn_cast<PHINode>(SI->getTrueValue()))
NonPhi = SI->getFalseValue();
else if (OrigPhi == dyn_cast<PHINode>(SI->getFalseValue()))
@@ -965,6 +964,9 @@ bool RecurrenceDescriptor::isFixedOrderRecurrence(PHINode *Phi, Loop *TheLoop,
Phi->getBasicBlockIndex(Latch) < 0)
return false;
+ if (!Phi->getNumUses())
+ return false;
+
// Get the previous value. The previous value comes from the latch edge while
// the initial value comes from the preheader edge.
auto *Previous = dyn_cast<Instruction>(Phi->getIncomingValueForBlock(Latch));
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorizationLegality.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorizationLegality.cpp
index 9de49d1bcfeac..73d49cbda2943 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorizationLegality.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorizationLegality.cpp
@@ -881,6 +881,9 @@ bool LoopVectorizationLegality::canVectorizeInstrs() {
continue;
}
+ if (!Phi->getNumUses())
+ continue;
+
reportVectorizationFailure("Found an unidentified PHI",
"value that could not be identified as "
"reduction is used outside the loop",
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index f2a541f5167a0..4ceb9c073e480 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -8483,6 +8483,9 @@ VPRecipeBuilder::tryToCreateWidenRecipe(Instruction *Instr,
if ((Recipe = tryToOptimizeInductionPHI(Phi, Operands, Range)))
return Recipe;
+ if (!Phi->getNumUses())
+ return nullptr;
+
VPHeaderPHIRecipe *PhiRecipe = nullptr;
assert((Legal->isReductionVariable(Phi) ||
Legal->isFixedOrderRecurrence(Phi)) &&
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/loop-vectorization-factors.ll b/llvm/test/Transforms/LoopVectorize/AArch64/loop-vectorization-factors.ll
index f2f58f89eb95d..8c53334832c3f 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/loop-vectorization-factors.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/loop-vectorization-factors.ll
@@ -759,50 +759,76 @@ define void @add_phifail(ptr noalias nocapture readonly %p, ptr noalias nocaptur
; CHECK-SAME: (ptr noalias nocapture readonly [[P:%.*]], ptr noalias nocapture [[Q:%.*]], i32 [[LEN:%.*]]) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CMP8:%.*]] = icmp sgt i32 [[LEN]], 0
-; CHECK-NEXT: br i1 [[CMP8]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
-; CHECK: for.body.preheader:
+; CHECK-NEXT: br i1 [[CMP8]], label [[ITER_CHECK:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK: iter.check:
; CHECK-NEXT: [[TMP0:%.*]] = zext i32 [[LEN]] to i64
-; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 16
-; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 8
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
+; CHECK: vector.main.loop.iter.check:
+; CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[TMP0]], 16
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 16
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <16 x i32> [ <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP1]]
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
-; CHECK-NEXT: [[TMP4]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
-; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <16 x i32> [[VECTOR_RECUR]], <16 x i32> [[TMP4]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
-; CHECK-NEXT: [[TMP6:%.*]] = add nuw nsw <16 x i32> [[TMP4]], <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
-; CHECK-NEXT: [[TMP7:%.*]] = trunc <16 x i32> [[TMP6]] to <16 x i8>
-; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[TMP1]]
-; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
-; CHECK-NEXT: store <16 x i8> [[TMP7]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
+; CHECK-NEXT: [[TMP5:%.*]] = add nuw nsw <16 x i32> [[TMP4]], <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
+; CHECK-NEXT: [[TMP6:%.*]] = trunc <16 x i32> [[TMP5]] to <16 x i8>
+; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[TMP1]]
+; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
+; CHECK-NEXT: store <16 x i8> [[TMP6]], ptr [[TMP8]], align 1
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
-; CHECK-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
-; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP4]], i32 15
-; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
-; CHECK: scalar.ph:
-; CHECK-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
-; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
+; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
+; CHECK: vec.epilog.iter.check:
+; CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[TMP0]], [[N_VEC]]
+; CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 8
+; CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
+; CHECK: vec.epilog.ph:
+; CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
+; CHECK-NEXT: [[N_MOD_VF2:%.*]] = urem i64 [[TMP0]], 8
+; CHECK-NEXT: [[N_VEC3:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF2]]
+; CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
+; CHECK: vec.epilog.vector.body:
+; CHECK-NEXT: [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT7:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[INDEX5]], 0
+; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP10]]
+; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
+; CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <8 x i8>, ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = zext <8 x i8> [[WIDE_LOAD6]] to <8 x i32>
+; CHECK-NEXT: [[TMP14:%.*]] = add nuw nsw <8 x i32> [[TMP13]], <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
+; CHECK-NEXT: [[TMP15:%.*]] = trunc <8 x i32> [[TMP14]] to <8 x i8>
+; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[TMP10]]
+; CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
+; CHECK-NEXT: store <8 x i8> [[TMP15]], ptr [[TMP17]], align 1
+; CHECK-NEXT: [[INDEX_NEXT7]] = add nuw i64 [[INDEX5]], 8
+; CHECK-NEXT: [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT7]], [[N_VEC3]]
+; CHECK-NEXT: br i1 [[TMP18]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; CHECK: vec.epilog.middle.block:
+; CHECK-NEXT: [[CMP_N4:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC3]]
+; CHECK-NEXT: br i1 [[CMP_N4]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
+; CHECK: vec.epilog.scalar.ph:
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.cond.cleanup.loopexit:
; CHECK-NEXT: br label [[FOR_COND_CLEANUP]]
; CHECK: for.cond.cleanup:
; CHECK-NEXT: ret void
; CHECK: for.body:
-; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
-; CHECK-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[CONV:%.*]], [[FOR_BODY]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ]
+; CHECK-NEXT: [[A_PHI:%.*]] = phi i32 [ [[CONV:%.*]], [[FOR_BODY]] ], [ 0, [[VEC_EPILOG_SCALAR_PH]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDVARS_IV]]
-; CHECK-NEXT: [[TMP11:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
-; CHECK-NEXT: [[CONV]] = zext i8 [[TMP11]] to i32
+; CHECK-NEXT: [[TMP19:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
+; CHECK-NEXT: [[CONV]] = zext i8 [[TMP19]] to i32
; CHECK-NEXT: [[ADD:%.*]] = add nuw nsw i32 [[CONV]], 2
; CHECK-NEXT: [[CONV1:%.*]] = trunc i32 [[ADD]] to i8
; CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDVARS_IV]]
@@ -810,7 +836,7 @@ define void @add_phifail(ptr noalias nocapture readonly %p, ptr noalias nocaptur
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[LEN]]
-; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
;
entry:
%cmp8 = icmp sgt i32 %len, 0
@@ -868,7 +894,7 @@ define i8 @add_phifail2(ptr noalias nocapture readonly %p, ptr noalias nocapture
; CHECK-NEXT: store <16 x i8> [[TMP9]], ptr [[TMP11]], align 1
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
; CHECK-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; CHECK-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP6]], i32 15
@@ -895,7 +921,7 @@ define i8 @add_phifail2(ptr noalias nocapture readonly %p, ptr noalias nocapture
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[LEN]]
-; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
;
entry:
br label %for.body
More information about the llvm-commits
mailing list