This is the mail archive of the
mailing list for the GCC project.
RE: [RFC PATCH, x86_64] Use -mno-sse[,2] to fall back to x87 FP argument passing convention
- From: "Menezes, Evandro" <evandro dot menezes at amd dot com>
- To: "Roger Sayle" <roger at eyesopen dot com>, "Uros Bizjak" <ubizjak at gmail dot com>
- Cc: "Michael Matz" <matz at suse dot de>, "GCC Patches" <gcc-patches at gcc dot gnu dot org>
- Date: Tue, 10 Oct 2006 18:39:00 -0500
- Subject: RE: [RFC PATCH, x86_64] Use -mno-sse[,2] to fall back to x87 FP argument passing convention
> On an Opteron 248 @ 2.21GHz
> -O3 -ffast-math -mfpmath=387 2578 MWIPS
> -O3 -ffast-math 2213 MWIPS
Here's what I got running on an Athlon 64 @ 2200MHz with SUSE 10.0 (therefore with math routines optimized for x86-64):
-O3 -ffast-math -mfpmath=387 2593 MWIPS
-O3 -ffast-math 2895 MWIPS
-O3 -ffast-math -m32 2539 MWIPS
-O3 -ffast-math -m32 -mfpmath=sse -msse2 2029 MWIPS
These results hint at the problem lying in libm, not x87 vs. SSE: SSE math is about 12% faster on 64-bit code.
Here's a comparison between the optimized routines and those currently in GLIBC: http://sources.redhat.com/ml/libc-alpha/2006-05/msg00078.html.
I'll see if I can run RH (therefore without math routines optimized for x86-64) on the same system and get results for the same binaries tomorrow.
Evandro Menezes AMD Austin, TX