Computer Chess Club Archives


Search

Terms

Messages

Subject: Widely distributed parallel chess engine?

Author: Ian Osgood

Date: 15:03:47 07/13/98


The organization Distributed.Net is currently using thousands of clients,
running on ordinary PC's, to exhaustively check RC-64 and DES-II keys to decode
a message and win a contest.  There is also interest in using the same
methodology to create a widely distributed parallel chess engine.

However, I think that this is not likely to work.  The search time would be
dominated by the network lag.  The more the search is parallelized, the more lag
would enter into the overall search time, and the less benefit you would obtain
from standard alpha-beta search techniques.  Since the clients would be widely
distributed, you would also have throw away the global transposition table.

The only previous work I can recall that sounds similar is Schaeffers work with
Pheonix.  However that was using a small number of computers connected on an
Ethernet LAN.

Does anyone else know of other work along these lines?  Any other opinions on
using large numbers of loosely coupled computers as a strategy for parallelizing
chess search?

The organizer of the distributed chess engine effort is Remy de Ruysscher
(remy@cyberservices.com).  He is the author of the program Shannon, which has
played in the Dutch Computer Championships.  A list of other possible projects
is at http://www.distributed.net/projects.html

Ian Osgood
iano@teleport.com



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.