[llvm] [AMDGPU] Enable sinking of free vector ops that will be folded into their uses (PR #162580)

[Gheorghe-Teodor Bercea via llvm-commits]

doru1004 wrote:

> This is still too much code re-parsing out all of the free vector situations, which TTI already has the code for. Do not repeat all of this. Just check for free and is vector instruction

I followed your suggestion, please have a look and let me know if this is what you had in mind.

https://github.com/llvm/llvm-project/pull/162580