The analysis result of DA caches pointers to AA, SCEV, and LI, but it
never checks for their invalidation. Fix that.
Details
Details
Diff Detail
Diff Detail
- Repository
- rL LLVM
Event Timeline
Comment Actions
Some of D56386 seems to be mixed up in here I think?
From what I can see of the new code, the invalidate looks a lot like the that from AA, so looks sensible. Maybe leave this open a day or so for other timezones to check, but LGTM.