[all-commits] [llvm/llvm-project] 2d6286: AMDGPU: Rewrite VGPR MFMAs to AGPR when directly c...

[Matt Arsenault via All-commits]

  Branch: refs/heads/users/arsenm/amdgpu/handle-direct-copy-to-agpr-mfma-rewrite
  Home:   https://github.com/llvm/llvm-project
  Commit: 2d6286c01e76c7de3e95a2462a536ffd9a1f3724
      https://github.com/llvm/llvm-project/commit/2d6286c01e76c7de3e95a2462a536ffd9a1f3724
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2025-08-07 (Thu, 07 Aug 2025)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPURewriteAGPRCopyMFMA.cpp
    M llvm/test/CodeGen/AMDGPU/rewrite-vgpr-mfma-to-agpr.ll

  Log Message:
  -----------
  AMDGPU: Rewrite VGPR MFMAs to AGPR when directly copied to AGPR class

Previously we were specifically looking for AV_* class registers,
and checking if the physreg assignment. Handle the case where the
copy is to an AGPR in the first place. In hindsight it would have
been way easier to handle this first, and this makes writing tests
a lot easier for the mechanical transforms.



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