[llvm] [AMDGPU] Form VOPD3 pairs with pair-local literal moves (PR #223431)

[Jay Foad via llvm-commits]

https://github.com/jayfoad commented:

Is there a test that clearly demonstrates the benefit? Why is S_MOV + VOPD3 better than VALU + VALU?

https://github.com/llvm/llvm-project/pull/223431