[PATCH] D32867: Warn about archives with no symbol table

[Rui Ueyama via llvm-commits]

I hacked it up and measured the performance. Oddly enough, the overhead is
negative. https://reviews.llvm.org/D32881

On Thu, May 4, 2017 at 1:57 PM, Rafael Avila de Espindola <
rafael.espindola at gmail.com> wrote:

> Rui Ueyama via Phabricator <reviews at reviews.llvm.org> writes:
>
> > ruiu added a comment.
> >
> > It seems it is only me who wants it to be silent, but if we can make it
> virtually zero-overhead, do you still want to warn on it?
>
> I am not fully decided. I am pretty sure that we can implement the same
> logic as ar, so there will be no semantic problems.
>
> Having a symbol table in the archive is then an optimization. If there
> is a particular case where it is a significant optimization, then sure,
> I think we should warn. For now it doesn't look necessary.
>
> > Currently reading public symbols for LazyObject is serialized, but it
> should be easy to parallelize, so there's a large room for improvement.
>
> If anyone has the time, an interesting thing to profile is why the
> archive symbol table is not providing a measurable saving. Maybe our
> archive code is slow, maybe the entire area is too fast to show up.
>

I think that's the case. Our archive code is probably slow.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170504/efe781a9/attachment.html>