[all-commits] [llvm/llvm-project] 3409f0: [AMDGPU] Fix invalid 64-bit DPP formed from foldin...

Domenic Nutile via All-commits all-commits at lists.llvm.org
Wed Sep 9 08:25:56 PDT 2026


  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 3409f0fd2aea08b57b1ee0047cdec5c26acfa02e
      https://github.com/llvm/llvm-project/commit/3409f0fd2aea08b57b1ee0047cdec5c26acfa02e
  Author: Domenic Nutile <domenic.nutile at gmail.com>
  Date:   2026-09-09 (Wed, 09 Sep 2026)

  Changed paths:
    M llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp
    M llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
    M llvm/test/CodeGen/AMDGPU/dpp_combine_rev_opcode.ll

  Log Message:
  -----------
  [AMDGPU] Fix invalid 64-bit DPP formed from folding a 32-bit DPP mov (#222216)

GCNDPPCombine already rejects 64-bit DPP when the DPP move itself is
64-bit, but a 32-bit move folded into a source of a 64-bit instruction
was not checked. On targets without FeatureDPALU_DPP this formed
V_LSHLREV_B64_e64_dpp, whose tied old operand got the move's 32-bit
register class instead of the 64-bit destination class.



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