[llvm] [AMDGPU] Fix negative immediate offset for unbuffered smem loads (PR #89165)

[Matt Arsenault via llvm-commits]

================
@@ -1985,35 +1985,47 @@ bool AMDGPUDAGToDAGISel::SelectScratchSVAddr(SDNode *N, SDValue Addr,
 // offsets available on CI.
 bool AMDGPUDAGToDAGISel::SelectSMRDOffset(SDValue ByteOffsetNode,
                                           SDValue *SOffset, SDValue *Offset,
-                                          bool Imm32Only, bool IsBuffer) const {
+                                          bool Imm32Only, bool IsBuffer,
+                                          bool HasSOffset,
+                                          int64_t ImmOffset) const {
   assert((!SOffset || !Offset) &&
          "Cannot match both soffset and offset at the same time!");
 
   ConstantSDNode *C = dyn_cast<ConstantSDNode>(ByteOffsetNode);
   if (!C) {
     if (!SOffset)
       return false;
+    bool Changed = false;
----------------
arsenm wrote:

There shouldn't be any need to track a "Changed" state. Nothing is changing here, it's a series of preconditions that succeeds or not. The end case should just return true If the mode matched. 

https://github.com/llvm/llvm-project/pull/89165