[all-commits] [llvm/llvm-project] ce091d: [AMDGPU] Mark WMMA machine instructions as converg...

[Syadus Sefat via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: ce091da5df6c095585b9cda48843f4a0a4952b79
      https://github.com/llvm/llvm-project/commit/ce091da5df6c095585b9cda48843f4a0a4952b79
  Author: Syadus Sefat <42645939+mssefat at users.noreply.github.com>
  Date:   2025-11-04 (Tue, 04 Nov 2025)

  Changed paths:
    M llvm/lib/Target/AMDGPU/VOP3PInstructions.td
    A llvm/test/CodeGen/AMDGPU/wmma-gfx12-convergent.mir

  Log Message:
  -----------
  [AMDGPU] Mark WMMA machine instructions as convergent (#165602)

The WMMA MI(s) are missing the isConvergent flag. This causes incorrect
behavior in passes like machine-sink, where WMMA instructions get sunk
into divergent branches.

This patch fixes the issue by setting the isConvergent flag to 1 in the
VOP3PInstructions.td file.



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