[llvm] [X86] Prefer VPMOVMSKB for nonnegative byte comparison masks (PR #216884)

[Oscar Priego via llvm-commits]

Opriego wrote:

> Would this be better handled in X86CompressEVEX? It already handles a number of MOVMSK patterns - x86 lowering can then just focus on canonicalisation

I was thinking along the same lines, I agree this would be a cleaner and more general solution, I'll work on the required changes and update the PR accordingly.

https://github.com/llvm/llvm-project/pull/216884