[PATCH] D63629: [JumpThreading][PR42085] Fold constant TIs before calculating LoopHeaders

[Eli Friedman via Phabricator via llvm-commits]

efriedma added a comment.

> there is a possibility of a miscompile when threading an edge across the LH

How do we actually get a miscompile?  The reason we don't thread across loop headers is to promote later loop optimizations, not because threading across loop headers is unsound.  In general, we have to support irreducible CFGs, and I'm worried this fix is covering up a bug in our handling of such code.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D63629/new/

https://reviews.llvm.org/D63629