This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] Powerpc/trampline: consider __NO_FPRS__
- From: "Ryan S. Arnold" <ryan dot arnold at gmail dot com>
- To: "Joseph S. Myers" <joseph at codesourcery dot com>
- Cc: Sebastian Andrzej Siewior <sebastian at breakpoint dot cc>, libc-alpha at sourceware dot org
- Date: Wed, 31 Oct 2012 16:29:51 -0500
- Subject: Re: [PATCH] Powerpc/trampline: consider __NO_FPRS__
- References: <20100618083214.GB5851@Chamillionaire.breakpoint.cc><AANLkTimKnCKmuVc2yEHf0ZvDueNdDUMMPejznOOca7ud@mail.gmail.com><Pine.LNX.4.64.1210252107010.30538@digraph.polyomino.org.uk>
On Thu, Oct 25, 2012 at 4:12 PM, Joseph S. Myers
<joseph@codesourcery.com> wrote:
> On Fri, 18 Jun 2010, Ryan Arnold wrote:
>
>> The currently blessed method for achieving what you want is to clone
>> this code and move a copy with FPU support into the powerpc32/fpu/
>> directory and strip the FPR save out of the powerpc32/dl-trampoline.S
>> code. This, of course, introduces code bloat of asm files which is
>> ... undesirable.
>
> I'd like to propose again that we apply the originally proposed patch from
> 2010 <http://sourceware.org/ml/libc-alpha/2010-06/msg00012.html>, as-is.
>
> Using extra sysdeps files just complicates things unduly. Hard-float and
> soft-float have two incompatible ABIs; a hypothetical nofpu/ build with a
> hard-float compiler is simply not a supported use case for glibc.
> Compilers configured for soft-float have defined __NO_FPRS__ since GCC
> 4.1, which means that all versions supported for building glibc will
> define this macro under the required conditions. So I think the tests of
> __NO_FPRS__ here are the right approach for addressing this issue.
I'm fine with this now. I'll see if the patch still applies, and if
not, I'll forward port and commit. Joseph, were you able to test it
with a no-fp toolchain?
Ryan