[llvm-branch-commits] [clang] clang/AMDGPU: Fix workgroup size builtins for nonuniform work group sizes (PR #185098)

[Matt Arsenault via llvm-branch-commits]

arsenm wrote:

> Then I'd probably say that OpenCL handling should be opt-in rather than opt-out, because AFAIK non-uniform work groups require OpenCL specific handling

No, the conservatively correct option should always be the default. For C I'd say it's completely meaningless. Somebody should look into setting the default LangOpt for OpenMP though 

https://github.com/llvm/llvm-project/pull/185098