[all-commits] [llvm/llvm-project] 9f47d4: [VPlan] Fold removeRedundantCanonicalIVs into repl...

[Florian Hahn via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 9f47d4c230f3eb49949c722a972136f6e01d1958
      https://github.com/llvm/llvm-project/commit/9f47d4c230f3eb49949c722a972136f6e01d1958
  Author: Florian Hahn <flo at fhahn.com>
  Date:   2026-05-07 (Thu, 07 May 2026)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
    M llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-masked-accesses.ll
    M llvm/test/Transforms/LoopVectorize/ARM/mve-gather-scatter-tailpred.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/mask-index-type.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/strided-accesses.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cast-intrinsics.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-fixed-order-recurrence.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-interleave.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll
    M llvm/test/Transforms/LoopVectorize/VPlan/first-order-recurrence-sink-replicate-region.ll
    M llvm/test/Transforms/LoopVectorize/VPlan/icmp-uniforms.ll
    M llvm/test/Transforms/LoopVectorize/VPlan/vplan-print-after-all.ll
    M llvm/test/Transforms/LoopVectorize/VPlan/vplan-sink-scalars-and-merge.ll
    M llvm/test/Transforms/LoopVectorize/first-order-recurrence-tail-folding.ll
    M llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll
    M llvm/test/Transforms/LoopVectorize/strict-fadd-interleave-only.ll

  Log Message:
  -----------
  [VPlan] Fold removeRedundantCanonicalIVs into replaceWideCanIVs. (#195545)

Unify logic to replace VPWidenCanonicalIVRecipes with scalar-steps or
VPWidenIntOrFpInductionRecipe in replaceWideCanonicalIVWithWideIV. This
reduces the code a bit and running later has the benefit that we do the
replacement after the wide mask compare has been replaced by
active-lane-mask/EVL. This means we do not need to drop wrap flags in
some cases, as the wide IV is not used for the mask.

PR: https://github.com/llvm/llvm-project/pull/195545



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications