[libcxx-dev] [cfe-dev] libc++ is not using always_inline anymore!

David Blaikie via libcxx-dev libcxx-dev at lists.llvm.org
Tue Oct 30 15:00:36 PDT 2018


What are the new semantics? That this ABI stability guarantee is provided
by hiding the functions in each user so they can't be deduplicated with
anotehr user's copy? (what about other copies that are from the same build?
I guess even those won't get coalesced/collapsed together? Would that be
useful to support?)

I assume this doesn't change the defaults, but does it make it any easier
for users who don't need the ABI stability guarantee? (or was it already
easy/no change here?)

On Mon, Oct 29, 2018 at 2:41 PM Louis Dionne via cfe-dev <
cfe-dev at lists.llvm.org> wrote:

> Folks,
> Today, I pushed a change that disables the use of always_inline for
> visibility purposes in libc++. The change was reviewed as
> https://reviews.llvm.org/D52405. This patch is the culminating point of
> several discussions about libc++'s visibility annotations.
> This change should have a positive impact on code size for almost any
> application using the libc++ headers. For example, Eric Fiselier reported a
> 7% improvement in size for the Chrome binary after applying this patch. We
> are still in the process of gathering more data like this, but we expect
> similar gains to be possible in other applications. The debugging
> experience should also be much improved as we don't inline every function
> into its caller.
> However, one word of caution: even though we were very careful not to
> break the ABI or any of libc++'s users ABI, this patch changes things that
> have been in place in libc++ for a long time. It is possible that we
> uncover problems with it, although we hope that won't be the case. If you
> start seeing problems you think are caused by this change, please notify me
> right away so we can take a look.
> I'll post updates here as we're able to publish more data and observations
> about the results of this patch.
> Thanks,
> Louis
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/libcxx-dev/attachments/20181030/3d53335e/attachment.html>

More information about the libcxx-dev mailing list