[PATCH] D44180: Improve --warn-symbol-ordering.

Rafael Avila de Espindola via llvm-commits llvm-commits at lists.llvm.org
Wed Mar 7 09:15:54 PST 2018


Rui Ueyama via Phabricator via llvm-commits
<llvm-commits at lists.llvm.org> writes:

> ruiu added a comment.
>
> I don't want to do stuff that is too clever, and I'm OK in either warning on it and don't sort, or don't warn on it and do sort. I just want to fix the issue that we currently tell user that the section is ignored but actually be sorted.
>
> To keep things simple, I'd suggest we do warn on ICF'ed functions and skip them when we see them in the loop. Does this sound OK?

I think your original idea of not warning is better.

On the thread about call graph based sorting pcc reported a win by
taking ICF into consideration. If there are multiple symbols in a
section using the highest priority symbol as the priority of the section
seems like the best solution for me.

I would recommend going with something like your original patch and then
having an independent discussion on whether we should add a warning when
two symbols are in the same section and would result in different
priorities.

Cheers,
Rafael


More information about the llvm-commits mailing list