[all-commits] [llvm/llvm-project] 2305da: [NFC][SLP][AMDGPU] Precommit a cross-block fmul fa...

[Dmitry Sidorov via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 2305da864cb3b66e1fe857c3018bf7d94253bb8a
      https://github.com/llvm/llvm-project/commit/2305da864cb3b66e1fe857c3018bf7d94253bb8a
  Author: Dmitry Sidorov <Dmitry.Sidorov at amd.com>
  Date:   2026-09-26 (Sat, 26 Sep 2026)

  Changed paths:
    A llvm/test/Transforms/SLPVectorizer/AMDGPU/cross-block-fmul-fadd.ll

  Log Message:
  -----------
  [NFC][SLP][AMDGPU] Precommit a cross-block fmul fadd contraction test (#226609)

Two contract fmuls feed contract fadds in different successors. AMDGPU
sinks such an fmul into the block of its user and fuses the pair, so the
scalar form is one fma per path. Record the current behaviour, the pair
is kept scalar when the fmul is operand 0 of the fadd and paired when it
is operand 1.



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