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

[Guo Chen via llvm-commits]

================
@@ -165,4 +165,5 @@ MACHINE_FUNCTION_PASS("si-pre-allocate-wwm-regs", SIPreAllocateWWMRegsPass())
 MACHINE_FUNCTION_PASS("si-pre-emit-peephole", SIPreEmitPeepholePass())
 MACHINE_FUNCTION_PASS("si-shrink-instructions", SIShrinkInstructionsPass())
 MACHINE_FUNCTION_PASS("si-wqm", SIWholeQuadModePass())
+MACHINE_FUNCTION_PASS("si-post-ra-16bit-mov-folding", SIPostRA16BitMovFoldingPass())
----------------
broxigarchen wrote:

done

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