[llvm] [NVPTX] Add some basic folds for ADDRSPACECAST (PR #129157)

[Alex MacLean via llvm-commits]

================
@@ -5209,6 +5209,26 @@ PerformBUILD_VECTORCombine(SDNode *N, TargetLowering::DAGCombinerInfo &DCI) {
   return DAG.getNode(ISD::BITCAST, DL, VT, PRMT);
 }
 
+static SDValue combineADDRSPACECAST(SDNode *N,
----------------
AlexMaclean wrote:

I agree there aren't non-bug cases where this fold would be useful. The only case I can imagine would be identifying invalid casts in code that will never execute allowing us to simplify by propagating undef. 

> The only possibility seems to be casts between shared::cta and shared::cluster. PTX docs are not quite clear on that (or I didn't find the right place in the docs yet). If you can clarify whether casts between shared variants are legal, then we could restore folding of transitive casts via generic AS for them.

Even if this cast were legal, we'd need to go through the generic AS to represent this in PTX, so folding it here would likely just make our lives harder. Further we don't represent shared::cta and shared::cluster as distinct AS making this a moot point. 

https://github.com/llvm/llvm-project/pull/129157