[all-commits] [llvm/llvm-project] d343b2: [SLP]Try the wider store slice when the leading VF...

[Alexey Bataev via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: d343b26ef687cefe30e7f8072d6befa6f2e0829d
      https://github.com/llvm/llvm-project/commit/d343b26ef687cefe30e7f8072d6befa6f2e0829d
  Author: Alexey Bataev <a.bataev at outlook.com>
  Date:   2026-09-23 (Wed, 23 Sep 2026)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    M llvm/test/Transforms/SLPVectorizer/AArch64/extract-fmul-cost.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/store-chain-leading-slice-wider-vf.ll

  Log Message:
  -----------
  [SLP]Try the wider store slice when the leading VF=2 slice fails

With the register VF 2 the instruction count check may reject the
leading slice of a store chain.
Try the double-width slice at the same position first and keep the wider
slices for the rest of the chain once one is vectorized.

Fixes the remaining perf regression from #221717.

Reviewers: 

Pull Request: https://github.com/llvm/llvm-project/pull/225823



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