[VPlan] Rename VPReverseVectorPointerRecipe to VPVectorEndPointerRecipe. NFC #131086
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
…pe. NFC After llvm#128718 lands there will be two ways of performing a reversed widened memory access, either by performing a consecutive unit-stride access and a reverse, or a strided access with a negative stride. Even though both produce a reversed vector, only the former needs VPReverseVectorPointerRecipe which computes a pointer to the last element of the widened vector. A strided reverse still needs a pointer to the first element of each part so it will use VPVectorPointerRecipe. This renames VPReverseVectorPointerRecipe to VPVectorEndPointerRecipe to clarify that a reversed access may not necessarily need a pointer to the last element.
|
@llvm/pr-subscribers-backend-risc-v @llvm/pr-subscribers-vectorizers Author: Luke Lau (lukel97) ChangesAfter #128718 lands there will be two ways of performing a reversed widened memory access, either by performing a consecutive unit-stride access and a reverse, or a strided access with a negative stride. Even though both produce a reversed vector, only the former needs VPReverseVectorPointerRecipe which computes a pointer to the last element of the widened vector. A strided reverse still needs a pointer to the first element of each part so it will use VPVectorPointerRecipe. This renames VPReverseVectorPointerRecipe to VPVectorEndPointerRecipe to clarify that a reversed access may not necessarily need a pointer to the last element. Full diff: https://github.com/llvm/llvm-project/pull/131086.diff 8 Files Affected:
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index c402b16f53b84..30b7ee24ef621 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -4506,7 +4506,7 @@ static bool willGenerateVectors(VPlan &Plan, ElementCount VF,
case VPDef::VPInstructionSC:
case VPDef::VPCanonicalIVPHISC:
case VPDef::VPVectorPointerSC:
- case VPDef::VPReverseVectorPointerSC:
+ case VPDef::VPVectorEndPointerSC:
case VPDef::VPExpandSCEVSC:
case VPDef::VPEVLBasedIVPHISC:
case VPDef::VPPredInstPHISC:
@@ -8348,7 +8348,7 @@ VPRecipeBuilder::tryToWidenMemory(Instruction *I, ArrayRef<VPValue *> Operands,
(CM.foldTailByMasking() || !GEP || !GEP->isInBounds())
? GEPNoWrapFlags::none()
: GEPNoWrapFlags::inBounds();
- VectorPtr = new VPReverseVectorPointerRecipe(
+ VectorPtr = new VPVectorEndPointerRecipe(
Ptr, &Plan.getVF(), getLoadStoreType(I), Flags, I->getDebugLoc());
} else {
VectorPtr = new VPVectorPointerRecipe(Ptr, getLoadStoreType(I),
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index b1288c42b20f2..d4ebc9fd95f9e 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -516,7 +516,7 @@ class VPSingleDefRecipe : public VPRecipeBase, public VPValue {
case VPRecipeBase::VPReplicateSC:
case VPRecipeBase::VPScalarIVStepsSC:
case VPRecipeBase::VPVectorPointerSC:
- case VPRecipeBase::VPReverseVectorPointerSC:
+ case VPRecipeBase::VPVectorEndPointerSC:
case VPRecipeBase::VPWidenCallSC:
case VPRecipeBase::VPWidenCanonicalIVSC:
case VPRecipeBase::VPWidenCastSC:
@@ -714,7 +714,7 @@ class VPRecipeWithIRFlags : public VPSingleDefRecipe {
R->getVPDefID() == VPRecipeBase::VPWidenCastSC ||
R->getVPDefID() == VPRecipeBase::VPWidenIntrinsicSC ||
R->getVPDefID() == VPRecipeBase::VPReplicateSC ||
- R->getVPDefID() == VPRecipeBase::VPReverseVectorPointerSC ||
+ R->getVPDefID() == VPRecipeBase::VPVectorEndPointerSC ||
R->getVPDefID() == VPRecipeBase::VPVectorPointerSC;
}
@@ -1518,20 +1518,21 @@ class VPWidenGEPRecipe : public VPRecipeWithIRFlags {
}
};
-/// A recipe to compute the pointers for widened memory accesses of IndexTy
-/// in reverse order.
-class VPReverseVectorPointerRecipe : public VPRecipeWithIRFlags,
- public VPUnrollPartAccessor<2> {
+/// A recipe to compute a pointer to the last element of each part of a widened
+/// memory access for widened memory accesses of IndexedTy. Used for
+/// VPWidenMemoryRecipes that are reversed.
+class VPVectorEndPointerRecipe : public VPRecipeWithIRFlags,
+ public VPUnrollPartAccessor<2> {
Type *IndexedTy;
public:
- VPReverseVectorPointerRecipe(VPValue *Ptr, VPValue *VF, Type *IndexedTy,
- GEPNoWrapFlags GEPFlags, DebugLoc DL)
- : VPRecipeWithIRFlags(VPDef::VPReverseVectorPointerSC,
+ VPVectorEndPointerRecipe(VPValue *Ptr, VPValue *VF, Type *IndexedTy,
+ GEPNoWrapFlags GEPFlags, DebugLoc DL)
+ : VPRecipeWithIRFlags(VPDef::VPVectorEndPointerSC,
ArrayRef<VPValue *>({Ptr, VF}), GEPFlags, DL),
IndexedTy(IndexedTy) {}
- VP_CLASSOF_IMPL(VPDef::VPReverseVectorPointerSC)
+ VP_CLASSOF_IMPL(VPDef::VPVectorEndPointerSC)
VPValue *getVFValue() { return getOperand(1); }
const VPValue *getVFValue() const { return getOperand(1); }
@@ -1559,10 +1560,9 @@ class VPReverseVectorPointerRecipe : public VPRecipeWithIRFlags,
return true;
}
- VPReverseVectorPointerRecipe *clone() override {
- return new VPReverseVectorPointerRecipe(getOperand(0), getVFValue(),
- IndexedTy, getGEPNoWrapFlags(),
- getDebugLoc());
+ VPVectorEndPointerRecipe *clone() override {
+ return new VPVectorEndPointerRecipe(getOperand(0), getVFValue(), IndexedTy,
+ getGEPNoWrapFlags(), getDebugLoc());
}
#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
diff --git a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp
index 6f6875f0e5e0e..671bb8ddd4a10 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp
@@ -248,7 +248,7 @@ Type *VPTypeAnalysis::inferScalarType(const VPValue *V) {
[](const auto *R) { return R->getScalarType(); })
.Case<VPReductionRecipe, VPPredInstPHIRecipe, VPWidenPHIRecipe,
VPScalarIVStepsRecipe, VPWidenGEPRecipe, VPVectorPointerRecipe,
- VPReverseVectorPointerRecipe, VPWidenCanonicalIVRecipe,
+ VPVectorEndPointerRecipe, VPWidenCanonicalIVRecipe,
VPPartialReductionRecipe>([this](const VPRecipeBase *R) {
return inferScalarType(R->getOperand(0));
})
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index d154d54c37862..ce87eb9f841a1 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -149,7 +149,7 @@ bool VPRecipeBase::mayHaveSideEffects() const {
case VPDerivedIVSC:
case VPPredInstPHISC:
case VPScalarCastSC:
- case VPReverseVectorPointerSC:
+ case VPVectorEndPointerSC:
return false;
case VPInstructionSC:
return mayWriteToMemory();
@@ -2081,7 +2081,7 @@ static Type *getGEPIndexTy(bool IsScalable, bool IsReverse,
: Builder.getInt32Ty();
}
-void VPReverseVectorPointerRecipe::execute(VPTransformState &State) {
+void VPVectorEndPointerRecipe::execute(VPTransformState &State) {
auto &Builder = State.Builder;
State.setDebugLocFrom(getDebugLoc());
unsigned CurrentPart = getUnrollPart(*this);
@@ -2107,8 +2107,8 @@ void VPReverseVectorPointerRecipe::execute(VPTransformState &State) {
}
#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
-void VPReverseVectorPointerRecipe::print(raw_ostream &O, const Twine &Indent,
- VPSlotTracker &SlotTracker) const {
+void VPVectorEndPointerRecipe::print(raw_ostream &O, const Twine &Indent,
+ VPSlotTracker &SlotTracker) const {
O << Indent;
printAsOperand(O, SlotTracker);
O << " = reverse-vector-pointer";
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 7646350ca0ed2..8beb7d96bd77d 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1758,7 +1758,7 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
}
for (VPUser *U : to_vector(Plan.getVF().users())) {
- if (auto *R = dyn_cast<VPReverseVectorPointerRecipe>(U))
+ if (auto *R = dyn_cast<VPVectorEndPointerRecipe>(U))
R->setOperand(1, &EVL);
}
diff --git a/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp b/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp
index 3fb6a5be79ff0..a36c2aeb3da5c 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp
@@ -311,12 +311,12 @@ void UnrollState::unrollRecipeByUF(VPRecipeBase &R) {
// Add operand indicating the part to generate code for, to recipes still
// requiring it.
if (isa<VPScalarIVStepsRecipe, VPWidenCanonicalIVRecipe,
- VPVectorPointerRecipe, VPReverseVectorPointerRecipe>(Copy) ||
+ VPVectorPointerRecipe, VPVectorEndPointerRecipe>(Copy) ||
match(Copy, m_VPInstruction<VPInstruction::CanonicalIVIncrementForPart>(
m_VPValue())))
Copy->addOperand(getConstantVPV(Part));
- if (isa<VPVectorPointerRecipe, VPReverseVectorPointerRecipe>(R))
+ if (isa<VPVectorPointerRecipe, VPVectorEndPointerRecipe>(R))
Copy->setOperand(0, R.getOperand(0));
}
}
diff --git a/llvm/lib/Transforms/Vectorize/VPlanValue.h b/llvm/lib/Transforms/Vectorize/VPlanValue.h
index 0a59b137bbd79..d0fa62978c1b4 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanValue.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanValue.h
@@ -335,7 +335,7 @@ class VPDef {
VPScalarCastSC,
VPScalarIVStepsSC,
VPVectorPointerSC,
- VPReverseVectorPointerSC,
+ VPVectorEndPointerSC,
VPWidenCallSC,
VPWidenCanonicalIVSC,
VPWidenCastSC,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanVerifier.cpp b/llvm/lib/Transforms/Vectorize/VPlanVerifier.cpp
index 1b3b69ea6a13d..3fde3bc9f03d3 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanVerifier.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanVerifier.cpp
@@ -143,7 +143,7 @@ bool VPlanVerifier::verifyEVLRecipe(const VPInstruction &EVL) const {
})
.Case<VPWidenStoreEVLRecipe, VPReductionEVLRecipe>(
[&](const VPRecipeBase *S) { return VerifyEVLUse(*S, 2); })
- .Case<VPWidenLoadEVLRecipe, VPReverseVectorPointerRecipe,
+ .Case<VPWidenLoadEVLRecipe, VPVectorEndPointerRecipe,
VPScalarPHIRecipe>(
[&](const VPRecipeBase *R) { return VerifyEVLUse(*R, 1); })
.Case<VPScalarCastRecipe>(
|
fhahn
reviewed
Mar 13, 2025
…ze/VPVectorEndPointerRecipe
alexey-bataev
approved these changes
Mar 14, 2025
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.
After #128718 lands there will be two ways of performing a reversed widened memory access, either by performing a consecutive unit-stride access and a reverse, or a strided access with a negative stride.
Even though both produce a reversed vector, only the former needs VPReverseVectorPointerRecipe which computes a pointer to the last element of each part. A strided reverse still needs a pointer to the first element of each part so it will use VPVectorPointerRecipe.
This renames VPReverseVectorPointerRecipe to VPVectorEndPointerRecipe to clarify that a reversed access may not necessarily need a pointer to the last element.