[all-commits] [llvm/llvm-project] 6dd229: [AMDGPU] Restrict DPP combine from performing bad ...

Domenic Nutile via All-commits all-commits at lists.llvm.org
Tue Sep 8 09:07:33 PDT 2026


  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 6dd22919138af830711ed4fb5b792482139b8eb0
      https://github.com/llvm/llvm-project/commit/6dd22919138af830711ed4fb5b792482139b8eb0
  Author: Domenic Nutile <domenic.nutile at gmail.com>
  Date:   2026-09-08 (Tue, 08 Sep 2026)

  Changed paths:
    M llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp
    M llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
    M llvm/lib/Target/AMDGPU/SIInstrInfo.h
    M llvm/test/CodeGen/AMDGPU/dpp_combine.mir
    M llvm/test/CodeGen/AMDGPU/dpp_combine_gfx11.mir
    A llvm/test/CodeGen/AMDGPU/dpp_combine_rev_opcode.ll
    A llvm/test/CodeGen/AMDGPU/dpp_combine_rev_opcode.mir

  Log Message:
  -----------
  [AMDGPU] Restrict DPP combine from performing bad transformations when handling certain REV insts that use Src1 as the DPP operand (#216835)

This issue was originally discovered due to a benchmark failure in the
downstream. The REV subtractions have always worked this way, and the
REV shift operands worked this way before GFX11



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


More information about the All-commits mailing list