Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Storing the PV in a search routine

Author: Robert Hyatt

Date: 16:47:14 06/09/98

Go up one level in this thread


On June 09, 1998 at 17:04:05, Scott Gasch wrote:

>Thanks a lot, your description was excellent, I've implemented it and it
>works.
>As a followup, you say that the memcpy won't hurt performance much but
>I'm not
>so sure of that... in your experience, is it better to use a 2D array
>for the
>PV or linked lists?  With the prior you have the issue of memcpys
>whereas with
>the latter you have the overhead of a memcpy per ply but can then link
>the
>ply-1 list to the ply list very easily.  Of course then you have to deal
>with
>free calls...
>
>Thanks again for the help.
>
>Scott


it hardly ever happens.  So you could copy it bit by bit using a
washing machine microprocessor and it wouldn't slow you down one
bit...  check your search (assuming you are using a PVS derivative)
and note how many nodes you enter where beta != alpha+1.  It will
almost *never* happen...  which means you can not possibly need to
back up a PV thru such a node...



This page took 0 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.