[llvm] [AMDGPU] Eliminate unnecessary packing in wider f16 vectors for sdwa/opsel-able instruction (PR #137137)

[Vikash Gupta via llvm-commits]

vg0204 wrote:

@jayfoad , @arsenm , @Pierre-vh , it would be appreciated if you could put further comments on this, thanks!

https://github.com/llvm/llvm-project/pull/137137