This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Comparing gcc and ccc on alpha with SPEC2000
- To: Brad Lucier <lucier at math dot purdue dot edu>
- Subject: Re: Comparing gcc and ccc on alpha with SPEC2000
- From: Toon Moene <toon at moene dot indiv dot nluug dot nl>
- Date: Fri, 04 Feb 2000 00:25:13 +0100
- CC: gcc at gcc dot gnu dot org
- Organization: Moene Computational Physics, Maartensdijk, The Netherlands
- References: <200002031421.JAA26231@polya.math.purdue.edu>
Brad Lucier wrote:
> Here are SPEC2000 comparisons between gcc+g77 and ccc+fort on alpha
> on 500 MHz 21264 alpha DS20 clone, 2 GB ram, 4 MB cache, Red Hat 6.0,
> kernel 2.2.13, binutils-2.9.5.0.16.
> Executive summary: the mainline gcc doesn't do so badly, g77 often
> sucks eggs.
Or ... *their* Fortran compiler is much more advanced than their C
compiler:
> With gcc version 2.96 20000127 (experimental), options -mcpu=ev6 -O3
> -ffast-math, MATHLIBOPT = -lcpml -lm and
> fort -arch ev6 -fast -O5 ONESTEP
> The first columns contains gcc+g77 times, the second ccc+fort times.
> 177.mesa 1400 369 379 * 1400 406 345*
mesa is written in C and gets a 406/369 = 1.1 boost of gcc over ccc -
hmmmm ...
> 301.apsi 2600 1640 159 * 2600 1156 225*
This is what I would be interested in: apsi is a weather model derived
dispersion code. Gain of Compaq Fortran over GNU Fortran: 1640/1156 =
1.4. That's certainly enough to pay attention (the cutoff for me is at
1.1).
Cheers,
--
Toon Moene (toon@moene.indiv.nluug.nl)
Saturnushof 14, 3738 XG Maartensdijk, The Netherlands
Phone: +31 346 214290; Fax: +31 346 214286
GNU Fortran: http://gcc.gnu.org/onlinedocs/g77_news.html