Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: some compiler experiments

Author: Jon Dart

Date: 15:07:10 07/10/04

Go up one level in this thread


I did manage to get the IPO and profile-guided optimizations to
work (the Users' guide is a bit sketchy here and doesn't mention
you have to use xilink). So the latest Intel results are (no
inline ASM, /Qprof_use /fast for optimization):

        middlegame  endgame
AMD     491 knps    818 knps
P4      559 knps    919 knps

A little below VC++ .NET for the middlegame but better than anything
else on the endgame test.

--Jon

On July 10, 2004 at 07:00:06, Tom Likens wrote:

>
>Hey Jon,
>
>I believe enabling the Intel compiler "ipo" and "pgo" options will go a
>long way towards restablishing parity between the compilers.  In fact,
>the only real use I find these days for WAC, besides a quick sanity check,
>is to use it for profile-guided optimization.
>
>Oddly, enough I've also had better luck with the 7.1 version of the Intel
>compiler, (unlike most other people, who've seen better results with 8.x).
>So if you've the time and the inclination, you might give that a try as well.
>
>regards,
>--tom



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.