[llvm] [NVPTX] Cleanup and refactor atomic lowering (PR #133781)

[Artem Belevich via llvm-commits]

================
@@ -994,6 +994,7 @@ NVPTXTargetLowering::NVPTXTargetLowering(const NVPTXTargetMachine &TM,
 
   setOperationAction(ISD::ADDRSPACECAST, {MVT::i32, MVT::i64}, Custom);
 
+  setOperationAction(ISD::ATOMIC_LOAD_SUB, {MVT::i32, MVT::i64}, Expand);
----------------
Artem-B wrote:

This could use a comment that PTX only provides `atom.add`.

Do we need something similar for other types? Why are i32/i64 singled out?

https://github.com/llvm/llvm-project/pull/133781