NFC right now (since the only user is the unit test), but will be used
in a future change.
Details
Diff Detail
Event Timeline
include/llvm/Analysis/ValueTracking.h | ||
---|---|---|
425 | Hm, this interface seems likely problematic. How does this handle the case where poison can be introduced by two different sources? Do we need an iterator abstraction? Or can we reliably just return null for those? (If the later, please clarify in the comment.) | |
lib/Analysis/ValueTracking.cpp | ||
3917 | What prevents this induction variable from being part of a longer phi cycle? i.e. a indvar with conditional increments for instance? i.e. this shouldn't be an assert of the documentation needs updated. |
Hm, this interface seems likely problematic. How does this handle the case where poison can be introduced by two different sources? Do we need an iterator abstraction? Or can we reliably just return null for those?
(If the later, please clarify in the comment.)