This is the mail archive of the
mailing list for the GCC project.
Re: powerpc new PLT and GOT
- From: Matt Thomas <matt at 3am-software dot com>
- To: Alan Modra <amodra at bigpond dot net dot au>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Thu, 12 May 2005 09:54:38 -0700
- Subject: Re: powerpc new PLT and GOT
- References: <20050512160523.GD12174@bubble.grove.modra.org>
Alan Modra wrote:
> This is where I'm at with gcc support for the new powerpc-linux PLT/GOT
> layout (see http://sources.redhat.com/ml/binutils/2005-05/msg00391.html).
> I'm not asking for commit approval yet; That ought to wait until I've
> thrown together glibc support as well so this can all be tested properly,
> but what I have here seems to do the right thing. So I'm looking for
> comments like "That's the Wrong Way. You ought to ..."
Not everyone who uses ABI_V4 uses glibc. Two, one might try to use
the new compiler/binutils on an older system who's dynamic loader doesn't
support the new PLT/GOT mechanism.
> Some things I know need attention:
> a) Should the new -fpic PLT/GOT code support be enabled by default? The
> linker will continue to generate the old GOT/PLT layout until a new
> glibc is available, a consequence of a "bl got-4" used in the current
> crti.o. This is fortunate, and means we don't need to do a configure
> test on glibc to figure whether the new PLT/GOT code is safe to use.
> However, the new GOT pointer load sequence is larger, (but might be
> quicker) and new PLT calls always need the GOT pointer, so code
> increases a little in size.
As I said before, the issue isn't the linker but the dynamic loader.
Unless you do a configure time to see if the dynamic loader does
the right thing I don't see how you can enable it by default.
Matt Thomas email: firstname.lastname@example.org
3am Software Foundry www: http://3am-software.com/bio/matt/
Cupertino, CA disclaimer: I avow all knowledge of this message.