[llvm] AMDGPU: Fold frame indexes into s_or_b32 and s_and_b32 (PR #102345)

[Changpeng Fang via llvm-commits]

================
@@ -2803,7 +2803,8 @@ bool SIRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator MI,
     }
     case AMDGPU::S_ADD_I32:
     case AMDGPU::S_ADD_U32: {
-      // TODO: Handle s_or_b32, s_and_b32.
+    case AMDGPU::S_OR_B32: {
+      // TODO: Handle s_and_b32 for ptrmask
----------------
changpeng wrote:

I do not fully understand how can the Offset be folded into the other operand the same way as add:  
**OtherOp.setImm(OtherOp.getImm() + Offset);**

https://github.com/llvm/llvm-project/pull/102345