This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix -march=athlon-tbird bug WRT SSE
- From: Kelley Cook <kcook34 at ford dot com>
- To: "David O'Brien" <obrien at FreeBSD dot org>, GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Thu, 28 Jul 2005 13:18:43 -0400
- Subject: Re: [PATCH] Fix -march=athlon-tbird bug WRT SSE
- Hop-count: 1
- References: <20050728163059.GA23621@dragon.NUXI.org>
- Reply-to: kcook at gcc dot gnu dot org
> 2005-07-28 David O'Brien <obrien@FreeBSD.org>
* config/i386/i386.h: Athlon-tbird does not have SSE support.
Personally, I would think that completely dropping the "athlon-tbird", "athlon-4", "k8",
"athlon-mp", "opteron", and "athlon-fx" aliases would make more sense as they don't add
anything useful to GCC.
For a specific example the Athlon rev B. (aka Thunderbird) was just a process shrink of
the original Athlon that the doubled the cache size and came on a PGA instead of a slot
card. GCC doesn't use cache size in any of its calculations while clearly packaging and
trace size aren't relevant.
This would leave "athlon", "athlon-xp", "x86-64", and "athlon64" in the two family trees.
Athlon is the original instruction set.
Athlon-xp is Athlon + SSE
Athlon64 is the K8 chip, which is AMD's current implementation of x86-64 (which includes 3DNow!)
Finally "x86-64" remains the generic x86-64 which includes SSE2, yet neither SSE3 or 3DNow!, and
therefore runs on either AMD original or Intel's copied implementation of x86-64.
Note that there are no Xeon nor Celeron references on the Intel side.