[PATCH] D145301: Add more efficient vector bitcast for AArch64

[Markus Everling via Phabricator via llvm-commits]

Sp00ph added a comment.

I think in general it would be way more convenient if LLVM just stored a flag for every vector that remembers whether the vector is guaranteed to be a "mask vector" where all elements are either 0 or -1 (where the flag would be true for all i1 vectors and would be preserved by trunc or sext operations). Then we wouldn't need to manually search whether the vector is a comparison result or not. Adding such a flag seems way out of scope for this PR though, so I think your strategy seems good considering the current constraints.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D145301/new/

https://reviews.llvm.org/D145301