[lld] r250808 - [ELF2] - Lazy relocation support for x86_64.

[Rui Ueyama via llvm-commits]

On x86-64, it seems that we can save 8 bytes for each PLT entry. For a
program whose size is clang, for example, that could save probably 10 KB or
so. Is that important?

On Tue, Oct 20, 2015 at 10:14 AM, Rafael Espíndola <
rafael.espindola at gmail.com> wrote:

> > Does that worth to do now? That sounds like an experimental "good to
> have"
> > feature. We are probably not going to use that often. (If you want to
> > resolve all symbols, you can use LD_BIND_NOW.) I actually requested in
> code
> > review to not add new code but replace the existing code in order to
> keep it
> > simple. I'm not totally against the idea itself, but in general I'm
> leaning
> > toward to not do that.
>
> Had this been our first PLT implementation I would agree.
>
> Given that all the necessary code for an eager PLT implementation is
> fresh on svn, it seems that now is a point where it would be easy to
> add it. There is also a move towards doing eager .got evaluation for
> security reasons (fedora 23 always uses -z now for example). Once that
> is the case, a simpler .plt/.got are nice additions.
>
> Given that there is '-z now', how about '-z eager'?
>
> Cheers,
> Rafael
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151020/c926d4a8/attachment.html>