This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: glibc-2.3.6 build errors on sparc64
- From: Jakub Jelinek <jakub at redhat dot com>
- To: Bruno Haible <bruno at clisp dot org>
- Cc: libc-alpha at sources dot redhat dot com
- Date: Mon, 5 Dec 2005 13:43:26 +0100
- Subject: Re: glibc-2.3.6 build errors on sparc64
- References: <200512051309.50054.bruno@clisp.org>
- Reply-to: Jakub Jelinek <jakub at redhat dot com>
On Mon, Dec 05, 2005 at 01:09:50PM +0100, Bruno Haible wrote:
> So, apparently, in the default configurations of sparc64 tools, the "fzero"
> instruction is invalid. Therefore I think it's a bug if memset.S uses this
> instruction unconditionally. Since the information about v9 vs. v9a/v9b is
> not available during preprocessing of the .S file, IMO glibc's configure
> script should check for it, or it should use something like
>
> stx %g0, [%o0]
> ldd [%o0], %f0
>
> instead of
>
> fzero %f0
No. There is no Linux sparc64 port that runs on non-UltraSPARC-I+ ISA
CPUs (the only ones were Fujitsu HALs CPUs and they were never supported
in Linux and most probably are never going to be, never Fujitsu SPARC64
CPUs support the v9a insns).
So you should have configured your GCC to default to -mcpu=ultrasparc
or later by default.
Jakub