[PATCH] D16986: [LICM] Don't assert on volatile accesses
Philip Reames via llvm-commits
llvm-commits at lists.llvm.org
Mon Feb 22 15:16:58 PST 2016
reames added a comment.
In http://reviews.llvm.org/D16986#359180, @nlopes wrote:
> But then store forwarding happens and LICM realizes (or doesn't) that two of the alias sets that were said to be disjoint by TBAA in fact alias.
Can you clarify this point? AFAIK, LICM does not do store forwarding. How did store forwarding come into the picture at all?
Also, if we ever discover two pointers mustalias, then by definition they must be the same alias set. If not, we have constructed *invalid, and incorrect* alias sets. Oh, I think I see the problem. You're saying that they're both noalias and mustalias discovered through two different mechanism. Ouch, ouch, ouch.
More information about the llvm-commits