Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Wierd hash problems

Author: Dan Homan

Date: 11:24:54 08/20/99

Go up one level in this thread


On August 20, 1999 at 13:40:28, James Robertson wrote:

>On August 20, 1999 at 12:24:44, Jon Dart wrote:
>
>>Actually, the first thing I'd look for is the logic you have for
>>storing and retrieving hash values. If you do not always correctly
>>identify the node type (fail-high, fail-low, or in bounds) or
>>if you do not properly set the search bounds after a hash hit,
>>you will have problems like you describe. Having "almost right"
>>logic will appear to work for a while, but sooner or later you
>>will pull a bad value out of the hash table and you will be
>>hosed.
>>
>>Unlike many areas of computer chess, there is really only one
>>right way to do this. Look at Arasan, Crafty, or Gnuchess source
>>and do what they do (in Crafty, look at hash.c).
>>
>>--Jon
>
>I have tried to look at Arasan source several times and was always completely
>confused. Crafty is confusing because the hash stuff is split between 2 files.
>EXchess is of some help. The only program that is not confusing is JRCP, and it
>has too many bugs. :(

The currently released version of EXchess (2.51) has some small bugs
in hash handling.  I forget exactly what they are... it is basically
right, but even small bugs can be a big problem in the right
position.  One bug that I can remember gives the same
behavior as your program... i.e. the MATE score is not always
properly adjusted to the ply depth.  I can't remember the exact
cause, but I think it had something to do with trying to return a
MATE score when I only had a bound rather than an exact
score in the table.

The newer version (currently in progress) has some fixes/improvements
in this area and many others.

 - Dan

>
>James



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.