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

[via llvm-commits]

================
@@ -1860,7 +1905,7 @@ bool AMDGPUDAGToDAGISel::SelectScratchSVAddr(SDNode *N, SDValue Addr,
           CurDAG->getTargetConstant(RemainderOffset, SDLoc(), MVT::i32));
         VAddr = SDValue(VMov, 0);
----------------
ruiling wrote:

Can we just change the code to `isUInt<31>(RemainderOffset)`? I think this should work well in practice even if the hardware start to interpret it as signed.

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