[llvm] [AMDGPU][NFC] Pre-commit test for constant-offset named barrier signal_var (PR #205217)

[Chinmay Deshpande via llvm-commits]

chinmaydd wrote:

> I've got a related PR over in https://github.com/llvm/llvm-project/pull/191506 ?

Right, thanks for pointing that out. Are you actively working on it ? We need the fold downstream, so it might be better if we split out the M0-folding optimization from the NULL barrier object discussion, which seems to be the central point of debate. If you'd be okay to fast-track that, I can close this stack.

https://github.com/llvm/llvm-project/pull/205217