[VPlan] Replace simplifyKnownEVL by fold in simplifyRecipe (NFCI).#183392
Closed
fhahn wants to merge 1 commit into
Closed
[VPlan] Replace simplifyKnownEVL by fold in simplifyRecipe (NFCI).#183392fhahn wants to merge 1 commit into
fhahn wants to merge 1 commit into
Conversation
Move the logic to simplify known EVL to simplifyRecipe as fold once we have a single VF. It does not look like this is only triggers when folding EVL with constant arguments. It cannot be a complete constant fold, as it needs to access the function's vscale range. Needed to avoid regressions in combination with llvm#181252.
Member
|
@llvm/pr-subscribers-vectorizers @llvm/pr-subscribers-llvm-transforms Author: Florian Hahn (fhahn) ChangesMove the logic to simplify known EVL to simplifyRecipe as fold once we have a single VF. It does not look like this is only triggers when folding EVL with constant arguments. It cannot be a complete constant fold, as it needs to access the function's vscale range. Needed to avoid regressions in combination with Full diff: https://github.com/llvm/llvm-project/pull/183392.diff 1 Files Affected:
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index c16149303b180..180f7270a2d50 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -34,6 +34,7 @@
#include "llvm/Analysis/MemoryLocation.h"
#include "llvm/Analysis/ScalarEvolutionPatternMatch.h"
#include "llvm/Analysis/ScopedNoAliasAA.h"
+#include "llvm/Analysis/ValueTracking.h"
#include "llvm/Analysis/VectorUtils.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/MDBuilder.h"
@@ -1532,6 +1533,25 @@ static void simplifyRecipe(VPSingleDefRecipe *Def, VPTypeAnalysis &TypeInfo) {
match(Def, m_ComputeReductionResult(m_VPIRValue(IRV))))
return Def->replaceAllUsesWith(IRV);
+ // Fold EVL(const_avl) -> const_avl when const_avl is known to be <= VF *
+ // vscale_min.
+ const APInt *AVL;
+ if (match(Def, m_EVL(m_APInt(AVL))) && size(Plan->vectorFactors()) == 1) {
+ ElementCount VF = *Plan->vectorFactors().begin();
+ uint64_t MinVScale = 1;
+ if (VF.isScalable()) {
+ const Function *F =
+ Plan->getScalarHeader()->getIRBasicBlock()->getParent();
+ MinVScale =
+ getVScaleRange(F, /*BitWidth=*/64).getUnsignedMin().getZExtValue();
+ }
+ unsigned MaxLanes = VF.getKnownMinValue() * MinVScale;
+ if (AVL->ule(MaxLanes)) {
+ return Def->replaceAllUsesWith(
+ Plan->getConstantInt(AVL->zextOrTrunc(32)));
+ }
+ }
+
// Some simplifications can only be applied after unrolling. Perform them
// below.
if (!Plan->isUnrolled())
@@ -2166,35 +2186,6 @@ static bool simplifyBranchConditionForVFAndUF(VPlan &Plan, ElementCount BestVF,
return true;
}
-/// From the definition of llvm.experimental.get.vector.length,
-/// VPInstruction::ExplicitVectorLength(%AVL) = %AVL when %AVL <= VF.
-static bool simplifyKnownEVL(VPlan &Plan, ElementCount VF,
- PredicatedScalarEvolution &PSE) {
- for (VPBasicBlock *VPBB : VPBlockUtils::blocksOnly<VPBasicBlock>(
- vp_depth_first_deep(Plan.getEntry()))) {
- for (VPRecipeBase &R : *VPBB) {
- VPValue *AVL;
- if (!match(&R, m_EVL(m_VPValue(AVL))))
- continue;
-
- const SCEV *AVLSCEV = vputils::getSCEVExprForVPValue(AVL, PSE);
- if (isa<SCEVCouldNotCompute>(AVLSCEV))
- continue;
- ScalarEvolution &SE = *PSE.getSE();
- const SCEV *VFSCEV = SE.getElementCount(AVLSCEV->getType(), VF);
- if (!SE.isKnownPredicate(CmpInst::ICMP_ULE, AVLSCEV, VFSCEV))
- continue;
-
- VPValue *Trunc = VPBuilder(&R).createScalarZExtOrTrunc(
- AVL, Type::getInt32Ty(Plan.getContext()), AVLSCEV->getType(),
- R.getDebugLoc());
- R.getVPSingleValue()->replaceAllUsesWith(Trunc);
- return true;
- }
- }
- return false;
-}
-
void VPlanTransforms::optimizeForVFAndUF(VPlan &Plan, ElementCount BestVF,
unsigned BestUF,
PredicatedScalarEvolution &PSE) {
@@ -2204,7 +2195,6 @@ void VPlanTransforms::optimizeForVFAndUF(VPlan &Plan, ElementCount BestVF,
bool MadeChange = tryToReplaceALMWithWideALM(Plan, BestVF, BestUF);
MadeChange |= simplifyBranchConditionForVFAndUF(Plan, BestVF, BestUF, PSE);
MadeChange |= optimizeVectorInductionWidthForTCAndVFUF(Plan, BestVF, BestUF);
- MadeChange |= simplifyKnownEVL(Plan, BestVF, PSE);
if (MadeChange) {
Plan.setVF(BestVF);
|
Contributor
Author
|
Found a case where this is not NFC and we need SCEV to simplify. Will add a test and close this one |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Move the logic to simplify known EVL to simplifyRecipe as fold once we have a single VF.
It does not look like this is only triggers when folding EVL with constant arguments. It cannot be a complete constant fold, as it needs to access the function's vscale range.
Needed to avoid regressions in combination with
#181252.