[all-commits] [llvm/llvm-project] 126134: AMDGPU: Preserve carry-out dead flag in buildShrun...

[Matt Arsenault via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 126134200f56180ef0b5cdab9f627d043f8e918d
      https://github.com/llvm/llvm-project/commit/126134200f56180ef0b5cdab9f627d043f8e918d
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2026-09-23 (Wed, 23 Sep 2026)

  Changed paths:
    M llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
    M llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp
    A llvm/test/CodeGen/AMDGPU/fold-immediate-operand-shrink-dead-carry.mir
    A llvm/test/CodeGen/AMDGPU/shrink-dead-vcc-def.mir
    R llvm/test/CodeGen/AMDGPU/shrink-v-cmp-wave32-dead-vcc-lo.mir

  Log Message:
  -----------
  AMDGPU: Preserve carry-out dead flag in buildShrunkInst (#225798)

Shrinking from the e64 to the e32 form drops the explicit carry-out def
in favor of an implicit vcc def. If the original carry out had a dead
flag, preserve it in the shrunk instruction's implicit-def operand.

Co-Authored-By: Claude Opus 5 <noreply at anthropic.com>



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