[clang] [llvm] AMDGPU: Add llvm.amdgcn.s.wait.event intrinsic (PR #180170)

[Matt Arsenault via cfe-commits]

arsenm wrote:

> Is this something that we can't just change in instruction selection to keep it consistent?

Semantically reinterpreting sounds like a terrible idea 

https://github.com/llvm/llvm-project/pull/180170