[ConstantFolding] Fold deinterleave2 of any splat vector not just zeroinitializer #144144
Conversation
This file contains hidden or 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
…oinitializer While there remove an unnecessary dyn_cast from Constant to Constant. Reverse a branch condition into an early out to reduce nesting.
|
@llvm/pr-subscribers-llvm-transforms Author: Craig Topper (topperc) ChangesWhile there remove an unnecessary dyn_cast from Constant to Constant. Reverse a branch condition into an early out to reduce nesting. Full diff: https://github.com/llvm/llvm-project/pull/144144.diff 2 Files Affected:
diff --git a/llvm/lib/Analysis/ConstantFolding.cpp b/llvm/lib/Analysis/ConstantFolding.cpp
index 64a0f4641250c..2b7a438a9ef01 100644
--- a/llvm/lib/Analysis/ConstantFolding.cpp
+++ b/llvm/lib/Analysis/ConstantFolding.cpp
@@ -3990,31 +3990,30 @@ ConstantFoldStructCall(StringRef Name, Intrinsic::ID IntrinsicID,
return ConstantStruct::get(StTy, SinResult, CosResult);
}
case Intrinsic::vector_deinterleave2: {
- auto *Vec = dyn_cast<Constant>(Operands[0]);
- if (!Vec)
- return nullptr;
-
+ auto *Vec = Operands[0];
auto *VecTy = cast<VectorType>(Vec->getType());
- unsigned NumElements = VecTy->getElementCount().getKnownMinValue() / 2;
- if (isa<ConstantAggregateZero>(Vec)) {
- auto *HalfVecTy = VectorType::getHalfElementsVectorType(VecTy);
- return ConstantStruct::get(StTy, ConstantAggregateZero::get(HalfVecTy),
- ConstantAggregateZero::get(HalfVecTy));
+
+ if (auto *EltC = Vec->getSplatValue()) {
+ ElementCount HalfEC = VecTy->getElementCount().divideCoefficientBy(2);
+ auto *HalfVec = ConstantVector::getSplat(HalfEC, EltC);
+ return ConstantStruct::get(StTy, HalfVec, HalfVec);
}
- if (isa<FixedVectorType>(Vec->getType())) {
- SmallVector<Constant *, 4> Res0(NumElements), Res1(NumElements);
- for (unsigned I = 0; I < NumElements; ++I) {
- Constant *Elt0 = Vec->getAggregateElement(2 * I);
- Constant *Elt1 = Vec->getAggregateElement(2 * I + 1);
- if (!Elt0 || !Elt1)
- return nullptr;
- Res0[I] = Elt0;
- Res1[I] = Elt1;
- }
- return ConstantStruct::get(StTy, ConstantVector::get(Res0),
- ConstantVector::get(Res1));
+
+ if (!isa<FixedVectorType>(Vec->getType()))
+ return nullptr;
+
+ unsigned NumElements = VecTy->getElementCount().getFixedValue() / 2;
+ SmallVector<Constant *, 4> Res0(NumElements), Res1(NumElements);
+ for (unsigned I = 0; I < NumElements; ++I) {
+ Constant *Elt0 = Vec->getAggregateElement(2 * I);
+ Constant *Elt1 = Vec->getAggregateElement(2 * I + 1);
+ if (!Elt0 || !Elt1)
+ return nullptr;
+ Res0[I] = Elt0;
+ Res1[I] = Elt1;
}
- return nullptr;
+ return ConstantStruct::get(StTy, ConstantVector::get(Res0),
+ ConstantVector::get(Res1));
}
default:
// TODO: Constant folding of vector intrinsics that fall through here does
diff --git a/llvm/test/Transforms/InstSimplify/ConstProp/vector-calls.ll b/llvm/test/Transforms/InstSimplify/ConstProp/vector-calls.ll
index 9dbe3d4e50ee1..14543f339db5d 100644
--- a/llvm/test/Transforms/InstSimplify/ConstProp/vector-calls.ll
+++ b/llvm/test/Transforms/InstSimplify/ConstProp/vector-calls.ll
@@ -66,3 +66,19 @@ define {<vscale x 4 x i32>, <vscale x 4 x i32>} @fold_scalable_vector_deinterlea
%1 = call {<vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.vector.deinterleave2.v4i32.v8i32(<vscale x 8 x i32> zeroinitializer)
ret {<vscale x 4 x i32>, <vscale x 4 x i32>} %1
}
+
+define {<vscale x 4 x i32>, <vscale x 4 x i32>} @fold_scalable_vector_deinterleave2_splat() {
+; CHECK-LABEL: define { <vscale x 4 x i32>, <vscale x 4 x i32> } @fold_scalable_vector_deinterleave2_splat() {
+; CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } { <vscale x 4 x i32> splat (i32 1), <vscale x 4 x i32> splat (i32 1) }
+;
+ %1 = call {<vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.vector.deinterleave2.v4i32.v8i32(<vscale x 8 x i32> splat (i32 1))
+ ret {<vscale x 4 x i32>, <vscale x 4 x i32>} %1
+}
+
+define {<vscale x 4 x float>, <vscale x 4 x float>} @fold_scalable_vector_deinterleave2_splatfp() {
+; CHECK-LABEL: define { <vscale x 4 x float>, <vscale x 4 x float> } @fold_scalable_vector_deinterleave2_splatfp() {
+; CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } { <vscale x 4 x float> splat (float 1.000000e+00), <vscale x 4 x float> splat (float 1.000000e+00) }
+;
+ %1 = call {<vscale x 4 x float>, <vscale x 4 x float>} @llvm.vector.deinterleave2.v4f32.v8f32(<vscale x 8 x float> splat (float 1.0))
+ ret {<vscale x 4 x float>, <vscale x 4 x float>} %1
+}
|
|
@llvm/pr-subscribers-llvm-analysis Author: Craig Topper (topperc) ChangesWhile there remove an unnecessary dyn_cast from Constant to Constant. Reverse a branch condition into an early out to reduce nesting. Full diff: https://github.com/llvm/llvm-project/pull/144144.diff 2 Files Affected:
diff --git a/llvm/lib/Analysis/ConstantFolding.cpp b/llvm/lib/Analysis/ConstantFolding.cpp
index 64a0f4641250c..2b7a438a9ef01 100644
--- a/llvm/lib/Analysis/ConstantFolding.cpp
+++ b/llvm/lib/Analysis/ConstantFolding.cpp
@@ -3990,31 +3990,30 @@ ConstantFoldStructCall(StringRef Name, Intrinsic::ID IntrinsicID,
return ConstantStruct::get(StTy, SinResult, CosResult);
}
case Intrinsic::vector_deinterleave2: {
- auto *Vec = dyn_cast<Constant>(Operands[0]);
- if (!Vec)
- return nullptr;
-
+ auto *Vec = Operands[0];
auto *VecTy = cast<VectorType>(Vec->getType());
- unsigned NumElements = VecTy->getElementCount().getKnownMinValue() / 2;
- if (isa<ConstantAggregateZero>(Vec)) {
- auto *HalfVecTy = VectorType::getHalfElementsVectorType(VecTy);
- return ConstantStruct::get(StTy, ConstantAggregateZero::get(HalfVecTy),
- ConstantAggregateZero::get(HalfVecTy));
+
+ if (auto *EltC = Vec->getSplatValue()) {
+ ElementCount HalfEC = VecTy->getElementCount().divideCoefficientBy(2);
+ auto *HalfVec = ConstantVector::getSplat(HalfEC, EltC);
+ return ConstantStruct::get(StTy, HalfVec, HalfVec);
}
- if (isa<FixedVectorType>(Vec->getType())) {
- SmallVector<Constant *, 4> Res0(NumElements), Res1(NumElements);
- for (unsigned I = 0; I < NumElements; ++I) {
- Constant *Elt0 = Vec->getAggregateElement(2 * I);
- Constant *Elt1 = Vec->getAggregateElement(2 * I + 1);
- if (!Elt0 || !Elt1)
- return nullptr;
- Res0[I] = Elt0;
- Res1[I] = Elt1;
- }
- return ConstantStruct::get(StTy, ConstantVector::get(Res0),
- ConstantVector::get(Res1));
+
+ if (!isa<FixedVectorType>(Vec->getType()))
+ return nullptr;
+
+ unsigned NumElements = VecTy->getElementCount().getFixedValue() / 2;
+ SmallVector<Constant *, 4> Res0(NumElements), Res1(NumElements);
+ for (unsigned I = 0; I < NumElements; ++I) {
+ Constant *Elt0 = Vec->getAggregateElement(2 * I);
+ Constant *Elt1 = Vec->getAggregateElement(2 * I + 1);
+ if (!Elt0 || !Elt1)
+ return nullptr;
+ Res0[I] = Elt0;
+ Res1[I] = Elt1;
}
- return nullptr;
+ return ConstantStruct::get(StTy, ConstantVector::get(Res0),
+ ConstantVector::get(Res1));
}
default:
// TODO: Constant folding of vector intrinsics that fall through here does
diff --git a/llvm/test/Transforms/InstSimplify/ConstProp/vector-calls.ll b/llvm/test/Transforms/InstSimplify/ConstProp/vector-calls.ll
index 9dbe3d4e50ee1..14543f339db5d 100644
--- a/llvm/test/Transforms/InstSimplify/ConstProp/vector-calls.ll
+++ b/llvm/test/Transforms/InstSimplify/ConstProp/vector-calls.ll
@@ -66,3 +66,19 @@ define {<vscale x 4 x i32>, <vscale x 4 x i32>} @fold_scalable_vector_deinterlea
%1 = call {<vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.vector.deinterleave2.v4i32.v8i32(<vscale x 8 x i32> zeroinitializer)
ret {<vscale x 4 x i32>, <vscale x 4 x i32>} %1
}
+
+define {<vscale x 4 x i32>, <vscale x 4 x i32>} @fold_scalable_vector_deinterleave2_splat() {
+; CHECK-LABEL: define { <vscale x 4 x i32>, <vscale x 4 x i32> } @fold_scalable_vector_deinterleave2_splat() {
+; CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } { <vscale x 4 x i32> splat (i32 1), <vscale x 4 x i32> splat (i32 1) }
+;
+ %1 = call {<vscale x 4 x i32>, <vscale x 4 x i32>} @llvm.vector.deinterleave2.v4i32.v8i32(<vscale x 8 x i32> splat (i32 1))
+ ret {<vscale x 4 x i32>, <vscale x 4 x i32>} %1
+}
+
+define {<vscale x 4 x float>, <vscale x 4 x float>} @fold_scalable_vector_deinterleave2_splatfp() {
+; CHECK-LABEL: define { <vscale x 4 x float>, <vscale x 4 x float> } @fold_scalable_vector_deinterleave2_splatfp() {
+; CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } { <vscale x 4 x float> splat (float 1.000000e+00), <vscale x 4 x float> splat (float 1.000000e+00) }
+;
+ %1 = call {<vscale x 4 x float>, <vscale x 4 x float>} @llvm.vector.deinterleave2.v4f32.v8f32(<vscale x 8 x float> splat (float 1.0))
+ ret {<vscale x 4 x float>, <vscale x 4 x float>} %1
+}
|
nikic
approved these changes
Jun 13, 2025
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/20/builds/11777 Here is the relevant piece of the build log for the reference |
tomtor
pushed a commit
to tomtor/llvm-project
that referenced
this pull request
Jun 14, 2025
…oinitializer (llvm#144144) While there remove an unnecessary dyn_cast from Constant to Constant. Reverse a branch condition into an early out to reduce nesting.
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.
While there remove an unnecessary dyn_cast from Constant to Constant. Reverse a branch condition into an early out to reduce nesting.