[all-commits] [llvm/llvm-project] 12c4c0: [AMDGPU] Handle expert scheduling VA_VDST WAR haza...

[Jay Foad via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 12c4c0bf0269218c6ffdddae1c895f892a3e4e75
      https://github.com/llvm/llvm-project/commit/12c4c0bf0269218c6ffdddae1c895f892a3e4e75
  Author: Jay Foad <jay.foad at amd.com>
  Date:   2026-07-17 (Fri, 17 Jul 2026)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUHWEvents.cpp
    M llvm/lib/Target/AMDGPU/AMDGPUHWEvents.def
    M llvm/lib/Target/AMDGPU/AMDGPUHWEvents.h
    M llvm/lib/Target/AMDGPU/AMDGPUWaitcntUtils.cpp
    M llvm/lib/Target/AMDGPU/AMDGPUWaitcntUtils.h
    M llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
    M llvm/test/CodeGen/AMDGPU/expert_scheduling_gfx12.mir

  Log Message:
  -----------
  [AMDGPU] Handle expert scheduling VA_VDST WAR hazards (#201619)

Handle WAR hazards where a VALU reading a VGPR is followed by a VMEM
writing the same VGPR. It is possible, though not very common, that the
VMEM will complete first. To guard against this the compiler must insert
a wait on VA_VDST.



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