[all-commits] [llvm/llvm-project] 9146ac: [AMDGPU] Prefer a safe V_PERM_PK16 follower in the...

[hidekisaito via All-commits]

  Branch: refs/heads/users/hidekisaito/perm-pk16-hazard-softcost
  Home:   https://github.com/llvm/llvm-project
  Commit: 9146ac4124f49087b6dbbb4f47dbfa1b2d3ffcc8
      https://github.com/llvm/llvm-project/commit/9146ac4124f49087b6dbbb4f47dbfa1b2d3ffcc8
  Author: Hideki Saito <hidekido at amd.com>
  Date:   2026-09-28 (Mon, 28 Sep 2026)

  Changed paths:
    M llvm/lib/Target/AMDGPU/GCNHazardRecognizer.cpp
    M llvm/lib/Target/AMDGPU/GCNHazardRecognizer.h
    A llvm/test/CodeGen/AMDGPU/vperm-pk16-postmisched-hazard.mir
    A llvm/test/CodeGen/AMDGPU/vperm-pk16-sched-softcost.mir

  Log Message:
  -----------
  [AMDGPU] Prefer a safe V_PERM_PK16 follower in the scheduler (gfx1250/gfx1251)

Stacked on the post-RA V_PERM_PK16 hazard fixup. V_PERM_PK16 must be
immediately followed by a "safe" instruction (see
SIInstrInfo::isVPermPk16SafeInstr) or the post-RA fixup has to insert a
forced-EXEC V_NOP. Teach GCNHazardRecognizer to bias a safe follower into
the slot right after a V_PERM_PK16 so that V_NOP can be avoided.

Assisted-by: Opus 4.8 Medium



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