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

[Vikash Gupta via llvm-commits]

vg0204 wrote:

Hi @krzysz00 , @jayfoad , @arsenm , @cdevadas , @frederik-h. Its been more than 2 months without any further progress on it due to my dilemma of which approach to follow as discussed above. So, can we come to some conclusion how should I move forward with this as of now to solve the issue at core!

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