[all-commits] [llvm/llvm-project] 03a618: [AMDGPU] Insert V_NOP after V_PERM_PK16 (gfx1250 h...

[hidekisaito via All-commits]

  Branch: refs/heads/users/hidekisaito/perm-pk16-hazard
  Home:   https://github.com/llvm/llvm-project
  Commit: 03a61893ffd38b526fd7e761baa03d1decbd1829
      https://github.com/llvm/llvm-project/commit/03a61893ffd38b526fd7e761baa03d1decbd1829
  Author: Hideki Saito <hidekido at amd.com>
  Date:   2026-09-23 (Wed, 23 Sep 2026)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPU.td
    M llvm/lib/Target/AMDGPU/GCNHazardRecognizer.cpp
    M llvm/lib/Target/AMDGPU/GCNHazardRecognizer.h
    M llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
    M llvm/lib/Target/AMDGPU/SIInstrInfo.h
    M llvm/test/CodeGen/AMDGPU/llvm.amdgcn.perm.pk.ll
    A llvm/test/CodeGen/AMDGPU/vperm-pk16-exec-hazard.ll
    A llvm/test/CodeGen/AMDGPU/vperm-pk16-exec-hazard.mir

  Log Message:
  -----------
  [AMDGPU] Insert V_NOP after V_PERM_PK16 (gfx1250 hazard)

On gfx1250 the V_PERM_PK16 family (V_PERM_PK16_B4/B6/B8_U4) has a hazard:
the instruction must be immediately followed by a "safe" instruction that
issues on the pipe which clears the hazard. Insert V_NOP as needed.

Also updated hasUnwantedEffectsWhenEXECEmpty() to ensure V_PERM_PK16 and
the inserted V_NOP are under non-zero EXEC.

Fixes: ROCM-26041

Assisted-by: Opus 4.8 Medium



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