[all-commits] [llvm/llvm-project] e203da: clang: Simplify emission of uniform-work-group-siz...

[Matt Arsenault via All-commits]

  Branch: refs/heads/users/arsenm/clang/simplify-uniform-work-group-size-attr-emission
  Home:   https://github.com/llvm/llvm-project
  Commit: e203da79c27dc7e6212847a7ed978bc824db9479
      https://github.com/llvm/llvm-project/commit/e203da79c27dc7e6212847a7ed978bc824db9479
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2026-03-06 (Fri, 06 Mar 2026)

  Changed paths:
    M clang/include/clang/Options/Options.td
    M clang/lib/CodeGen/CGCall.cpp
    M clang/test/CodeGenCUDA/convergent.cu
    M clang/test/CodeGenCUDA/incomplete-func-ptr-type.cu
    M clang/test/CodeGenHIP/default-attributes.hip
    M clang/test/CodeGenHIP/hip_weak_alias.cpp
    M clang/test/CodeGenOpenCL/cl20-device-side-enqueue-attributes.cl
    M clang/test/CodeGenOpenCL/convergent.cl

  Log Message:
  -----------
  clang: Simplify emission of uniform-work-group-size attribute

This wasn't strictly using the LangOpt field for this property,
and hardcoding the OpenCL version. It was also going out of its
way to specifically annotate specific calling conventions. Just
unconditionally emit it on all functions. The uniform work group
assumption must hold in the entire module, not just the entrypoint.
This also theoretically saves work for the attributor propagation.

This will avoid the need to repeat the same logic in builtin function
codegen.



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