Mailinglist Archive: opensuse-amd64 (380 mails)

< Previous Next >
Re: [suse-amd64] Hardware support
  • From: Bjorn Tore Sund <bjornts@xxxxxxxxx>
  • Date: Mon, 2 Feb 2004 09:11:54 +0000 (UTC)
  • Message-id: <Pine.LNX.4.58.0402020958280.22020@xxxxxxxxxxxxxxx>

On Mon, 2 Feb 2004, Odon Farkas wrote:

> > Hmm, I would be interested in some example code showing such behaviour.
> > Because on AMD64 (in 64bit mode) we have twice as many registers and use
> > SSE2 by default. This means that in theory (and mostly in practice too)
> > floating point on AMD64 should outperform anything on 32bit quite a bit.
> > This has nothing to do with 64bit or not, but rather with the much better
> > instruction set available in 64bit mode.
> >
> I was also surprised, since I expected the same. Gaussian uses the PGI
> fortran compiler and the code was linked with AMD's ACML. It is very
> likely a current handicap of the PGI compiler. Also a possible reason is
> that the cache can store more 32 bit integers than the default 64-bit
> ones when compiled with i8. Gaussian is not open source (see
>, so I am not allowed to give example code, the whole
> program is cca. 1M lines in Fortran. I could not find any SpecFP
> benchmark which was done on Opteron/Athlon64 in 64/bit mode and could
> outperform the same machine in 32-bit mode. However, I really hope this
> will happen...

This is surprising. We've been using the PGI 64 bit compilers on Opteron
since this summer with very good results. Haven't tried Gaussian, but
various self-written mathematical (including floating point) applications
have run with the expected speed. SLES 8 for AMD64 and PGI 5.0, now up
to PGI 5.1. With this said, we've mostly been using the C++ compiler.

I would contemplate the possibility of a weakness in the Gaussian code,

Bjørn Tore Sund Phone: (+47) 555-84894 Stupidity is like a
System administrator Fax: (+47) 555-89672 fractal; universal and
Math. Department Mobile: (+47) 918 68075 infinitely repetitive.
University of Bergen VIP: 81724
teknisk@xxxxxxxxx Email: bjornts@xxxxxxxxx

< Previous Next >
Follow Ups