[llvm] [NVPTX] Pull invariant load identification into IR pass (PR #138015)

[Artem Belevich via llvm-commits]

================
@@ -766,46 +766,12 @@ NVPTX::Scope NVPTXDAGToDAGISel::getOperationScope(MemSDNode *N,
   llvm_unreachable("unhandled ordering");
 }
 
-static bool canLowerToLDG(MemSDNode *N, const NVPTXSubtarget &Subtarget,
-                          unsigned CodeAddrSpace, MachineFunction *F) {
+static bool canLowerToLDG(const MemSDNode *N, const NVPTXSubtarget &Subtarget,
----------------
Artem-B wrote:

May as well make N a constref.

https://github.com/llvm/llvm-project/pull/138015