[llvm] SCEV: teach isImpliedViaOperations about samesign (PR #124270)

[Ramkumar Ramachandra via llvm-commits]

artagnon wrote:

> Yeah, there is a general problem with LFTR being too aggressive. We might want to disable it for the case where it just wants to canonicalize the predicate. Introducing extra instructions for the trip count calculation in that case doesn't seem worthwhile (unlike for example the case where we can replace an and/or of multiple conditions with a single one).

If I understand correctly, this is quite non-trivial, as we don't immediately have information about whether the LFTR is just canonicalizing a predicate or performing additional simplifications as well.

https://github.com/llvm/llvm-project/pull/124270