[llvm] [VPlan] Cost truncated widened inductions via ::computeCost. (PR #212786)
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Thu Jul 30 02:13:54 PDT 2026
================
@@ -5672,26 +5675,23 @@ LoopVectorizationPlanner::precomputeCosts(VPlan &Plan, ElementCount VF,
auto TC = getSmallConstantTripCount(PSE.getSE(), OrigLoop);
bool IsFullyUnrolled = TC == VF && !Plan.hasTailFolded();
SmallPtrSet<const Value *, 4> WidenedIVs;
- bool HasTruncatedIV = false;
if (IsFullyUnrolled) {
addFullyUnrolledInstructionsToIgnore(OrigLoop, Legal->getInductionVars(),
CostCtx.SkipCostComputation);
} else {
// Inductions represented by a VPWidenIntOrFpInductionRecipe have their cost
// computed by the recipe, so collect their phis to skip the legacy
- // increment cost below. If any induction is truncated the VPlan-based cost
- // will diverge. Still fall back to the legacy cost model for now.
- VPRegionBlock *LoopRegion = Plan.getVectorLoopRegion();
- for (VPRecipeBase &R : *LoopRegion->getEntryBasicBlock())
- if (auto *WideIV = dyn_cast<VPWidenIntOrFpInductionRecipe>(&R)) {
- HasTruncatedIV |= WideIV->getTruncInst() != nullptr;
- if (PHINode *IVPhi = WideIV->getPHINode())
- WidenedIVs.insert(IVPhi);
- }
+ // increment cost below.
+ if (VPRegionBlock *LoopRegion = Plan.getVectorLoopRegion())
----------------
fhahn wrote:
This change is not needed and just come back after a rebase mistake on my end after landing https://github.com/llvm/llvm-project/pull/202232 which had the simplificaiton. Gone now, thanks
https://github.com/llvm/llvm-project/pull/212786
More information about the llvm-commits
mailing list