[llvm] [AMDGPU] merge 16bit mov pairs in post-RA peephole (PR #208625)

[Jay Foad via llvm-commits]

================
@@ -44,6 +44,9 @@ class SIPreEmitPeephole {
   void updateMLIBeforeRemovingEdge(MachineBasicBlock *From,
                                    MachineBasicBlock *To) const;
   bool optimizeSetGPR(MachineInstr &First, MachineInstr &MI) const;
+  bool mergeSingleMovB16Pair(MachineInstr &Lo, MachineInstr &Hi,
+                             bool IsHiFirst) const;
+  bool mergeMovB16Pair(MachineFunction &MF) const;
----------------
jayfoad wrote:

Nit
```suggestion
  bool mergeMovB16Pairs(MachineFunction &MF) const;
```

https://github.com/llvm/llvm-project/pull/208625