[llvm] [AMDGPU] Folding imm offset in more cases for scratch access (PR #70634)

[Jay Foad via llvm-commits]

================
@@ -1839,6 +1883,7 @@ bool AMDGPUDAGToDAGISel::SelectScratchSVAddr(SDNode *N, SDValue Addr,
   int64_t ImmOffset = 0;
 
   SDValue LHS, RHS;
+  SDValue FullAddr = Addr;
----------------
jayfoad wrote:

I would prefer to use `Addr` for the full address and `Base` for the part without the immediate offset.

https://github.com/llvm/llvm-project/pull/70634