[all-commits] [llvm/llvm-project] a00f7b: [SLP]Use poor-throughput instructions as vectoriza...

[Alexey Bataev via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: a00f7be392d9468779af59d86bed12d17aa578c0
      https://github.com/llvm/llvm-project/commit/a00f7be392d9468779af59d86bed12d17aa578c0
  Author: Alexey Bataev <a.bataev at outlook.com>
  Date:   2026-07-07 (Tue, 07 Jul 2026)

  Changed paths:
    M llvm/include/llvm/Transforms/Vectorize/SLPVectorizer.h
    M llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    M llvm/test/Transforms/PhaseOrdering/X86/vector-reductions.ll
    M llvm/test/Transforms/SLPVectorizer/X86/c-ray.ll
    M llvm/test/Transforms/SLPVectorizer/X86/delayed-gather-emission.ll
    M llvm/test/Transforms/SLPVectorizer/X86/poor-throughput-seeds.ll
    M llvm/test/Transforms/SLPVectorizer/X86/reduction2.ll
    M llvm/test/Transforms/SLPVectorizer/X86/vec3-reorder-reshuffle.ll

  Log Message:
  -----------
  [SLP]Use poor-throughput instructions as vectorization seeds

Seed vectorization from expensive poor-throughput ops (fdiv/frem/fsqrt
and target-expensive calls) even when they do not feed a store or reduction.

Compile time effect https://llvm-compile-time-tracker.com/compare.php?from=8cdf6346f46c505928a9fb9d3ef9e8ce125a2108&to=27a3f0a5c0489c91bf386e480289fdcc9dc9c8b7&stat=instructions:u

These compile time regressions are the side effects of the vectorization
vectorization gains and cannot be reduced further (note Bullet).

Fixes #38780

Reviewers: hiraditya, bababuck, RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/206518



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