Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Some questions about large hash tables

Author: William Penn

Date: 10:19:16 10/11/03

Go up one level in this thread


On October 11, 2003 at 03:57:57, Peter Stayne wrote:

>Lately, I've been having Deep Fritz 7 analyze positions for long periods of time
>(from 10-24 hours). I'm certainly going to up my RAM soon from 1GB to 2GB, and
>was curious as to what, exactly, will be the benefits for chess programs in
>these overlong sessions.
>
>I guess I'm mostly curious what a chess program does when it runs out of hash
>space. Is there costly reorganization of the positions in the hash space to fit
>new positions in? Or, once it's filled, does it just keep whatever's in there
>and work completely from the CPU? Or does it depend on the engine? Would I
>notice a speedup in reaching the higher ply numbers (20-ply+)?
>
>Thanks,
>
>Pete

I don't have an answer for you, but am also curious to know more in this area,
so will be watching for further responses.

I'll just mention that I have 1GB RAM installed, maximum for my computer, and
couldn't find any reliable (tested/proven) reference information about proper
sizing or use of larger hash tables. So I devised some tests for 1-4 hour
runtimes in Infinite Analysis mode, attempting to answer some of those basic
questions about hash, but was unsuccessful. I was using the Shredder704.eng
engine. The main problem was that I couldn't reproduce a particular "depth of
search" nor the same (identical) line of analysis. Even if I ran it a particular
test position twice using identical starting conditions and settings, it
invariably produced different results. So no baseline was possible, nothing was
comparable, and I quit trying. Perhaps your Deep Fritz engine gives more
reproducible results so you could run your own tests(?).
WP



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.