[llvm] AMDGPU: Drop and upgrade llvm.amdgcn.atomic.csub/cond.sub to atomicrmw (PR #105553)

[via llvm-commits]

anjenner wrote:

> > Could you be more specific about what tests are needed? If you could point me to the corresponding tests for one of the other atomicrmw operations (so that I can copy them and change the operation to usub_cond/usub_sat), that would be ideal.
> 
> Pretty much everything in test/Transforms/AtomicExpand. expand-atomic-_-agent.ll and expand-atomic-_-system.ll are the most comprehensive tests

I copied what already existed for the inc and dec atomicrmw operations, so added tests to expand-atomic-{i8,i16}{-system,}.ll as well as expand-atomic-{i32,i64}-{agent,system}.ll in PR https://github.com/llvm/llvm-project/pull/141068 . Hope that looks reasonable.

https://github.com/llvm/llvm-project/pull/105553