[llvm] [AMDGPU] Waterfall loop codegen improvement in SIInstrInfo (PR #192415)

[Jay Foad via llvm-commits]

jayfoad wrote:

> > > I have just pushed a tentative fix: add `_term` versions of `V_CMPX_EQ_*` Machine IR instructions and use them in the current PR.
> > 
> > 
> > Maybe we don't even need non-terminator versions of V_CMPX instructions, since they always write to EXEC?
> 
> Are you suggesting a change to all `V_CMPX*` instruction definitions not specific to this PR? It is causing some test failures, probably minor issues. Would you like me to make a separate PR for it?

Yes that's what I'm suggesting. I have no idea how easy or difficult it would be. Please feel free to try it if you want!

https://github.com/llvm/llvm-project/pull/192415