[SLP][REVEC] Make shufflevector can be vectorized with ReorderIndices and ReuseShuffleIndices. #114965
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@llvm/pr-subscribers-llvm-transforms Author: Han-Kuan Chen (HanKuanChen) ChangesFull diff: https://github.com/llvm/llvm-project/pull/114965.diff 2 Files Affected:
diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 427b8bd0e75ab0..b14465ecc1dc91 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -15374,9 +15374,6 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
case Instruction::ShuffleVector: {
Value *V;
if (SLPReVec && !E->isAltShuffle()) {
- assert(E->ReuseShuffleIndices.empty() &&
- "Not support ReuseShuffleIndices yet.");
- assert(E->ReorderIndices.empty() && "Not support ReorderIndices yet.");
setInsertPointAfterBundle(E);
Value *Src = vectorizeOperand(E, 0, PostponedPHIs);
if (E->VectorizedValue) {
@@ -15394,6 +15391,9 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
[&SVSrc](int Mask) { return SVSrc->getShuffleMask()[Mask]; });
V = Builder.CreateShuffleVector(SVSrc->getOperand(0), NewMask);
propagateIRFlags(V, E->Scalars, VL0);
+ if (auto *I = dyn_cast<Instruction>(V))
+ V = propagateMetadata(I, E->Scalars);
+ V = FinalShuffle(V, E);
} else {
assert(E->isAltShuffle() &&
((Instruction::isBinaryOp(E->getOpcode()) &&
@@ -15524,11 +15524,11 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
transformScalarShuffleIndiciesToVector(VecTy->getNumElements(), Mask);
}
V = Builder.CreateShuffleVector(V0, V1, Mask);
- }
- if (auto *I = dyn_cast<Instruction>(V)) {
- V = propagateMetadata(I, E->Scalars);
- GatherShuffleExtractSeq.insert(I);
- CSEBlocks.insert(I->getParent());
+ if (auto *I = dyn_cast<Instruction>(V)) {
+ V = propagateMetadata(I, E->Scalars);
+ GatherShuffleExtractSeq.insert(I);
+ CSEBlocks.insert(I->getParent());
+ }
}
E->VectorizedValue = V;
diff --git a/llvm/test/Transforms/SLPVectorizer/revec-shufflevector.ll b/llvm/test/Transforms/SLPVectorizer/revec-shufflevector.ll
index 1fc0b0306d1194..39ba9d52d11e9b 100644
--- a/llvm/test/Transforms/SLPVectorizer/revec-shufflevector.ll
+++ b/llvm/test/Transforms/SLPVectorizer/revec-shufflevector.ll
@@ -82,3 +82,40 @@ entry:
store <4 x i32> %4, ptr %8, align 4
ret void
}
+
+define void @test4(ptr %in) {
+; CHECK-LABEL: @test4(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br label [[LABEL0:%.*]]
+; CHECK: label0:
+; CHECK-NEXT: br label [[LABEL2:%.*]]
+; CHECK: label1:
+; CHECK-NEXT: br label [[LABEL2]]
+; CHECK: label2:
+; CHECK-NEXT: br label [[LABEL0]]
+;
+entry:
+ br label %label0
+
+label0:
+ %0 = load <8 x float>, ptr %in, align 4
+ %1 = shufflevector <8 x float> %0, <8 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %2 = shufflevector <8 x float> %0, <8 x float> zeroinitializer, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %1, <4 x float> zeroinitializer, <4 x float> zeroinitializer)
+ %4 = call <4 x float> @llvm.fma.v4f32(<4 x float> %2, <4 x float> zeroinitializer, <4 x float> zeroinitializer)
+ %5 = call <4 x float> @llvm.fma.v4f32(<4 x float> %1, <4 x float> zeroinitializer, <4 x float> zeroinitializer)
+ %6 = call <4 x float> @llvm.fma.v4f32(<4 x float> %2, <4 x float> zeroinitializer, <4 x float> zeroinitializer)
+ br label %label2
+
+label1:
+ br label %label2
+
+label2:
+ %7 = phi <4 x float> [ %3, %label0 ], [ zeroinitializer, %label1 ]
+ %8 = phi <4 x float> [ %4, %label0 ], [ zeroinitializer, %label1 ]
+ %9 = phi <4 x float> [ %5, %label0 ], [ zeroinitializer, %label1 ]
+ %10 = phi <4 x float> [ %6, %label0 ], [ zeroinitializer, %label1 ]
+ br label %label0
+}
+
+declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>)
|
|
@llvm/pr-subscribers-vectorizers Author: Han-Kuan Chen (HanKuanChen) ChangesFull diff: https://github.com/llvm/llvm-project/pull/114965.diff 2 Files Affected:
diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 427b8bd0e75ab0..b14465ecc1dc91 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -15374,9 +15374,6 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
case Instruction::ShuffleVector: {
Value *V;
if (SLPReVec && !E->isAltShuffle()) {
- assert(E->ReuseShuffleIndices.empty() &&
- "Not support ReuseShuffleIndices yet.");
- assert(E->ReorderIndices.empty() && "Not support ReorderIndices yet.");
setInsertPointAfterBundle(E);
Value *Src = vectorizeOperand(E, 0, PostponedPHIs);
if (E->VectorizedValue) {
@@ -15394,6 +15391,9 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
[&SVSrc](int Mask) { return SVSrc->getShuffleMask()[Mask]; });
V = Builder.CreateShuffleVector(SVSrc->getOperand(0), NewMask);
propagateIRFlags(V, E->Scalars, VL0);
+ if (auto *I = dyn_cast<Instruction>(V))
+ V = propagateMetadata(I, E->Scalars);
+ V = FinalShuffle(V, E);
} else {
assert(E->isAltShuffle() &&
((Instruction::isBinaryOp(E->getOpcode()) &&
@@ -15524,11 +15524,11 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
transformScalarShuffleIndiciesToVector(VecTy->getNumElements(), Mask);
}
V = Builder.CreateShuffleVector(V0, V1, Mask);
- }
- if (auto *I = dyn_cast<Instruction>(V)) {
- V = propagateMetadata(I, E->Scalars);
- GatherShuffleExtractSeq.insert(I);
- CSEBlocks.insert(I->getParent());
+ if (auto *I = dyn_cast<Instruction>(V)) {
+ V = propagateMetadata(I, E->Scalars);
+ GatherShuffleExtractSeq.insert(I);
+ CSEBlocks.insert(I->getParent());
+ }
}
E->VectorizedValue = V;
diff --git a/llvm/test/Transforms/SLPVectorizer/revec-shufflevector.ll b/llvm/test/Transforms/SLPVectorizer/revec-shufflevector.ll
index 1fc0b0306d1194..39ba9d52d11e9b 100644
--- a/llvm/test/Transforms/SLPVectorizer/revec-shufflevector.ll
+++ b/llvm/test/Transforms/SLPVectorizer/revec-shufflevector.ll
@@ -82,3 +82,40 @@ entry:
store <4 x i32> %4, ptr %8, align 4
ret void
}
+
+define void @test4(ptr %in) {
+; CHECK-LABEL: @test4(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br label [[LABEL0:%.*]]
+; CHECK: label0:
+; CHECK-NEXT: br label [[LABEL2:%.*]]
+; CHECK: label1:
+; CHECK-NEXT: br label [[LABEL2]]
+; CHECK: label2:
+; CHECK-NEXT: br label [[LABEL0]]
+;
+entry:
+ br label %label0
+
+label0:
+ %0 = load <8 x float>, ptr %in, align 4
+ %1 = shufflevector <8 x float> %0, <8 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %2 = shufflevector <8 x float> %0, <8 x float> zeroinitializer, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %1, <4 x float> zeroinitializer, <4 x float> zeroinitializer)
+ %4 = call <4 x float> @llvm.fma.v4f32(<4 x float> %2, <4 x float> zeroinitializer, <4 x float> zeroinitializer)
+ %5 = call <4 x float> @llvm.fma.v4f32(<4 x float> %1, <4 x float> zeroinitializer, <4 x float> zeroinitializer)
+ %6 = call <4 x float> @llvm.fma.v4f32(<4 x float> %2, <4 x float> zeroinitializer, <4 x float> zeroinitializer)
+ br label %label2
+
+label1:
+ br label %label2
+
+label2:
+ %7 = phi <4 x float> [ %3, %label0 ], [ zeroinitializer, %label1 ]
+ %8 = phi <4 x float> [ %4, %label0 ], [ zeroinitializer, %label1 ]
+ %9 = phi <4 x float> [ %5, %label0 ], [ zeroinitializer, %label1 ]
+ %10 = phi <4 x float> [ %6, %label0 ], [ zeroinitializer, %label1 ]
+ br label %label0
+}
+
+declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>)
|
and ReuseShuffleIndices.
HanKuanChen
force-pushed
the
slp-revec-shufflevector-FinalShuffle
branch
from
5e0653e to
b3ecc2a
Compare
alexey-bataev
approved these changes
Nov 6, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.