Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A test position for chess programs(importance of tablebases)

Author: Dann Corbit

Date: 11:30:57 01/19/01

Go up one level in this thread


On January 19, 2001 at 14:13:24, Dieter Buerssner wrote:

>>I have nothing to tell you I'm afraid, but I would like you to tell *me*
>>something :) My program slows to a crawl in a soup of tablebase accesses!
>>Yours seems to be pretty speedy even though you said it also slows down.
>>Mine approaches 10 kN/s iso the usual 200 kN/s if I remember correctly ...
>>
>>What's the trick? My program goes recursively down the tree, and the first
>>thing it does at every ply is check the tb's if there are 5 pieces or less.
>>If that succeeds it returns with the proper score. Otherwise it goes on
>>with the usual hash lookup etc.
>>
>>Is there anything you can help me with (other than 'look at the source' :)?
>>I'd like to see the mate 39/40 too :)
>
>I do store EGTB scores in the hash tables. They are specially flagged
>and stored as exact scores. I first probe the hash table, and then EGTBs.
>This will avoid many disk accesses. Also, I ignore the depth stored
>in the hash, for EGTB positions, because the information will be valid
>for any depth.
>
>Nevertheless, sometimes I only get about 20% of the normal speed, when
>very many TB accesses are done.

Give the tablebase files 100 megs of hash and you will see the performance go
back up.



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.