[PATCH] D35887: isKnownNonNull: recognize GEP inbounds as non-null

[Nuno Lopes via Phabricator via llvm-commits]

nlopes abandoned this revision.
nlopes added a comment.

In https://reviews.llvm.org/D35887#823029, @efriedma wrote:

> See https://reviews.llvm.org/D28026


Ah, thanks for the pointer.
Then I'll abandon this patch. I'm not sure why I understand why we have isKnownNotNull and isKnownNotZero.  The difference seems to be that the latter is recursive and the former is not.  I dunno if that was intended or not.
The current situation is that not all tricks are duplicated in the two functions, so the clients miss out if they use isKnownNotNull().
Any opinions or pointers to historical artifacts?


Repository:
  rL LLVM

https://reviews.llvm.org/D35887