Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Efficient hash algorithm?

Author: Pat King

Date: 05:01:42 09/23/98

Go up one level in this thread



On September 22, 1998 at 12:24:37, Roberto Waldteufel wrote:
>On September 22, 1998 at 08:14:30, Pat King wrote:
>>On September 22, 1998 at 05:52:43, Roberto Waldteufel wrote:
>>
[big snip]
>
>This seems fine, but it does require a new variable "new.ply" to be carted
>around the search, otherwise move extensions (eg after checking moves) would
>cause problems: the depth parameter would not give a reliable indication of true
>depth. Possibly it might be more efficient to allow some innaccuracy in order to
>avoid this, eg replace if:
>
>1) gamelength>oldgamelength+offset
>2)depth>olddepth
>
>Here the gamelength refers only to the moves that have actually been played, not
>moves in the search. It is therefore constant over any one search, although it
>increases for searches done throughout the game. The offset could be set at some
>value slightly greater than the usual depth achieved in the searches, which
>would of course depend on the time controls and the nature of the position.
>
>Since the gamelength is held in a global variable anyway, this method would not
>require any additional data to be updated during the search, yet would still
>overwrite the outdated entries. What do you think?
>
>Best wishes,
>Roberto
The way my program is structured, new.ply is readily available as is.
current.ply is global
and equivalent to your gamelength. So I think we are pretty much in agreement,
subject
to our programs' individual differences.

Pat



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.