Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Debugging hash tables

Author: Anthony Cozzie

Date: 07:39:51 03/13/04

Go up one level in this thread


On March 13, 2004 at 07:30:24, Andrew Wagner wrote:

>Well, my program seems to have sufficient eval, but is always getting
>out-searched (it only gets 4-6 ply in 15 0!). So I finally decided to implement
>hash tables. I thoroughly tested all my hashing in my move-generation code, so I
>am at least 99% certain that is correct. However, my evals are all over the
>place. Some very strange scores, like +- 5 pawns in an even material position.
>Anybody have some suggestions as to what hash statistics to produce and look at?
>I don't know anything about hash table statistics. Thanks!

I will give you the same advice Vincent gave me.

Write a log of everything your engine does, and use grep to search it.  Give
each node an id.

example:

$ grep "000000001" trueno_log.txt
000000001 trans_probe: true [-15, UL]
000000001 searching child 000000001 [quiescence], move Nxb3
000000001 child returned: -17
....

It is not that hard to walk down the tree.  I eventually wrote a graphical
viewer and a binary format, but this will do for a start.  You will be amazed at
 how much stupid stuff you are doing.

anthony





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.