[llvm] [AMDGPU] Only treat unconditional branches as implicit XCNT drains (PR #219206)

[Madhur Kumar via llvm-commits]

MadhurKumar004 wrote:

Sorry for the delayed response.
> But there is still a bug, right? 

Yes there is still a bug. I closed this because i didn't had clear understanding of the issue and that the fix is still very hacky & conservative because it adds wait even where it isn't required. Would you still want me to reopen?

https://github.com/llvm/llvm-project/pull/219206