[llvm-branch-commits] [clang] clang: Simplify OpenMP triple adjustment (PR #189265)

[Joseph Huber via llvm-branch-commits]

jhuber6 wrote:

> I did break this in the middle of working on this, there's only the one test

Quite tempted to just deprecate this "feature". The defaults make sense because they match the default CUDA / HIP architectures, but this is primarily for 'convenience' when we already have a much more convenient option for OpenMP. Would likely mean old code gets warnings about incompatible triples, not sure what `amdgcn--` would mean in OpenMP contexts.

https://github.com/llvm/llvm-project/pull/189265