[PATCH] D85460: [SVE] No need to AND(X, 0x1) when truncating a scalable mask

[Eli Friedman via Phabricator via llvm-commits]

efriedma added a comment.

The semantics of TRUNCATE is that it throws away the high bits.  For most types, this happens implicitly, but we have to be a little more explicit here.   We can't assume a the operand to a random TRUNCATE to i1 is produced by a compare operation.

If it turns out the operand to the AND is actually an all-ones/all-zeros value, we can optimize out the AND, sure.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D85460/new/

https://reviews.llvm.org/D85460