[llvm] [AMDGPU] (xor (cmp_eq x, 1), -1) -> cmp_neq x, 1 (PR #133698)

[Pierre van Houtryve via llvm-commits]

Pierre-vh wrote:

Ah, I didn't pay enough attention, I thought the pattern was selecting a cmp but it's a trunc.
The title is a bit confusing, it should say we're selecting `(xor (i1 (trunc x)), -1)` into a cmp_ne` or something similar.

https://github.com/llvm/llvm-project/pull/133698