Author: Robert Hyatt
Date: 20:29:42 12/27/02
Go up one level in this thread
On December 27, 2002 at 18:20:22, Bruce Moreland wrote: >On December 27, 2002 at 16:51:14, Robert Hyatt wrote: > >>On December 27, 2002 at 02:52:43, Bruce Moreland wrote: >> >>>On December 27, 2002 at 02:06:25, Uri Blass wrote: >>> >>>>No >>>>It is not always correct. >>>> >>>>Tscp believes that the position is the same if the empty squares are the same >>>>In most of the games it cause no errors but there are cases when it can cause >>>>stupid blunders. >>> >>>I suggest that the best way to do this is to make a small hash table and insert >>>hash keys of previously seen positions into it when a node is entered, and >>>remove them with the node is left. >>> >>>The table does not need to be large, since depth of search is not infinite. A >>>small table won't blow cache. >>> >>>It can be large enough that collisions are rare. Collisions are handled by >>>moving to the next element, anyway. >>> >>>The method works in multiprocessor implementationss with some overhead but >>>without causing horrific bugs. >> >>What are you doing to prevent a processor searching one pathway of the >>tree from seeing a repetition position from a processor searching a >>completely different path in the tree? > >I use one table per processor. > >bruce OK... basically what I do with the repetition list...
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.