[llvm] AMDGPU: Mark wqm_vote as SourceOfDivergence (PR #188831)

[Jay Foad via llvm-commits]

https://github.com/jayfoad requested changes to this pull request.

The i1 result of this intrinsic is uniform if the i1 input is uniform. It shouldn't need any special handling.

> It looks like it selects directly to WQM, so why isn't it just named WQM?

There is an s_wqm intrinsic too. It looks like the only difference is that that one is defined on uniform i32/i64 values instead of i1 per-lane values.

https://github.com/llvm/llvm-project/pull/188831