[llvm] [AMDGPU] frame index elimination hit assertion for scavenged nonreg (PR #130287)

[Matt Arsenault via llvm-commits]

================
@@ -2802,18 +2805,20 @@ bool SIRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator MI,
           DstReg = TmpReg;
         }
 
-        auto AddI32 = BuildMI(*MBB, *MI, DL, MI->getDesc())
-                          .addDef(DstReg, RegState::Renamable)
-                          .addReg(MaterializedReg, RegState::Kill)
-                          .add(OtherOp);
-        if (DeadSCC)
-          AddI32.setOperandDead(3);
+        if (TmpReg) {
----------------
arsenm wrote:

The condition here still doesn't make sense. The TmpReg is not used below. I'm confused by the whole flow now, it would make more sense to check MaterializedReg 

https://github.com/llvm/llvm-project/pull/130287