This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] powerpc64-linux bi-arch support
- From: Michael Meissner <gcc-pat at the-meissners dot org>
- To: gcc-patches at gcc dot gnu dot org
- Date: Mon, 2 Jun 2003 18:37:25 -0400
- Subject: Re: [PATCH] powerpc64-linux bi-arch support
- References: <20030531135635.O13397@devserv.devel.redhat.com> <200306022208.SAA22070@makai.watson.ibm.com>
On Mon, Jun 02, 2003 at 06:08:25PM -0400, David Edelsohn wrote:
> >>>>> Jakub Jelinek writes:
>
> Jakub> But then either linux.h would need to define RS6000_CALL_GLUE the same way,
> Jakub> or -mcall-aixdesc should be disallowed for powerpc-*-linux* and
> Jakub> powerpc64-*-linux* targets.
> Jakub> If linux.h inherits the cror definition from sysv.h and linux64.h defines
> Jakub> it to nop unconditionally, then suddenly --target powerpc-ibm-linux and
> Jakub> --target powerpc64-ibm-linux --with-cpu=default32 behave differently,
> Jakub> which is IMHO not desirable. The only difference between those two should
> Jakub> be that the latter supports -m64 while the former does not.
>
> I cannot see how -mcall-aixdesc makes sense for powerpc*-*-linux*
It doesn't for Linux. It is or was needed for some of Cygnus's (now Red Hat)
embedded customers who started with GCC before I added the System V/eabi
calling sequences (roughly 1994-1995 time frame), and they needed to be able to
use the same calling sequence they were used to (basically AIX without the 3
word function pointer if memory serves). I have no idea if those customers are
still using PowerPC platforms or not, or whether they are still using this
option. To simplify the number of multilibs that were built, I restricted it
to just big endian, and I imagine a similar restriction to 32-bit would allow
the current users to use it, but encourage any new development to use the
official eabi calling sequence.
--
Michael Meissner
email: gnu@the-meissners.org
http://www.the-meissners.org