[PATCH] D96881: Avoid Bitcast-GEP fusion for void pointers

[Nikita Popov via Phabricator via llvm-commits]

nikic added a comment.

I'm not really sure what to do here. I see @lebedev.ri's point that ideally phi translation would be capable of detecting that an "equivalent" GEP exists, but this doesn't really fit into the current model of the phi translation analysis, so this seems like a pragmatic workaround.

The real solution here of course is to remove both pointer element types and GEP types and thus obviate the problem, but that's still far away.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D96881/new/

https://reviews.llvm.org/D96881