[llvm] [AMDGPU] Fix canonicalization of truncated values. (PR #83054)

[Matt Arsenault via llvm-commits]

arsenm wrote:

> Awesome! I wonder if we could do something similar for is_divergent and is_uniform, so we don't need separate DivergentUnaryFrag/DivergentBinFrag/etc.

Probably not. I'm running into a few issues trying to clean up the other uses. First, the predicate seems to not actually get applied when this is used as the source of a complex pattern. Plus we need at least another variant with an operand for the non-leaf cases 

https://github.com/llvm/llvm-project/pull/83054