[all-commits] [llvm/llvm-project] a8ccbb: [SLP]Do not propagate reduction fast-math flags to...

[Alexey Bataev via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: a8ccbb407250d6c83a7bf2a33f3cd0d2f0096424
      https://github.com/llvm/llvm-project/commit/a8ccbb407250d6c83a7bf2a33f3cd0d2f0096424
  Author: Alexey Bataev <a.bataev at outlook.com>
  Date:   2026-09-22 (Tue, 22 Sep 2026)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    M llvm/test/Transforms/SLPVectorizer/AArch64/loop-accumulator-reduction.ll
    M llvm/test/Transforms/SLPVectorizer/X86/loop-accumulator-reduction.ll

  Log Message:
  -----------
  [SLP]Do not propagate reduction fast-math flags to the loop bypass select

The select between the final loop accumulator reduction and the values
bypassing the loop inherited the reduction fast-math flags from the
builder, so a NaN or Inf bypass value became poison under nnan/ninf.

Fixes https://github.com/llvm/llvm-project/pull/221598#issuecomment-5782195781

Reviewers: 

Pull Request: https://github.com/llvm/llvm-project/pull/225532



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