[PATCH] D81911: [IR] Fix getBaseObject for GlobalAlias-to-GlobalIFunc

[Teresa Johnson via Phabricator via llvm-commits]

tejohnson added a comment.

In D81911#2226172 <https://reviews.llvm.org/D81911#2226172>, @nextsilicon-itay-bookstein wrote:

> @tejohnson would simply not emitting an AliasSummary in computeAliasSummary when getBaseObject() returns nullptr be sufficient, then?

We need a full reference graph in the thin link otherwise we may do incorrect optimizations (e.g. think that the aliasee is dead). I think then you'll run into the problem that D82745 <https://reviews.llvm.org/D82745> is trying to address (ifunc removed because we don't summarize it).

Perhaps the right fix for now is to implement @DmitryPolukhin 's suggestion of returning GI there, so at least we get the alias summary correct. Then when D82745 <https://reviews.llvm.org/D82745> is completed, the IFunc itself will be summarized.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D81911/new/

https://reviews.llvm.org/D81911