[PATCH] D29966: [SelectionDAG] Try to recompute LiveOutInfo of PHI

[Hal Finkel via Phabricator via llvm-commits]

hfinkel added a comment.

Do you have a test case?

It looks like you're making this algorithm directly recursive with a worst-case recursion depth given only by the number of blocks. Is the right? Can we say something about the worst case here?


https://reviews.llvm.org/D29966