Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: GCC 3.1 fastest compiler at K7 !!

Author: Dieter Buerssner

Date: 13:49:11 05/19/02

Go up one level in this thread


On May 18, 2002 at 18:47:31, Vincent Diepeveen wrote:

I tried gcc 3.1 as well. After some bad experience with former versions, I
decided to not download a rpm (which could easily after installing leave me with
a compiler I don't want to use, and the old one is gone), but rather build it
myself, with the option to install it in a path of its own. Most went well, but
make install had a (probably nonsignificant) error. Some shell line was just not
working (it had to do with "CATALOGS"). After some little hacking I got it
installed.

But my results were a bit dissapointing. With gcc 2.9.3 I allways had the best
results (not only with the chess engine) by using -O only. So I tried those same
options. And Yace was running slower 5% slower. Then I tried -O2 (which got me a
slower executable with 293), and it was about 1% faster in average than with
293. I tired -O3, and it was about 10% slower than the -O2 compile.

Then, with exactly one position (the start position), I tried the
-fprofile-arcs, followed by -fbranch-predictions. Speed up of about 4%. All in
all, it is faster than the older gcc, but not too impressive.

Regards,
Dieter



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.