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

[Krzysztof Drewniak via llvm-commits]

https://github.com/krzysz00 commented:

I'm going to leave the rest of the review to people who understand this infrastructure better - I don't have a good enough sense to know if the core of this transformation is how it needs to be. But I can say I like the results I'm seeing in the tests

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