Author: Vincent Diepeveen
Date: 06:53:36 07/19/99
Go up one level in this thread
On July 18, 1999 at 13:05:23, Francis Monkman wrote: > >On July 18, 1999 at 12:56:35, Robert Hyatt wrote: > >>The idea sounds attractive, until you realize that the game tree search is >>an exponential problem, not a linear one..... That makes it _very_ difficult >>for such a task to be done on computers... > >But why not treat the computers as nodes in a tree, with sub-delegation >software? (Just a thought -- but I built a working parallel audio synthesizer >out of multiple TMS99000s back in '84, so I've been 'thinking parallel' for a >while -- though not in chess. That's why I thought of you. Hope you don't mind) > >Say the machines are in a pool. Starting from root, one machine picks the next n >(=number of legal moves) machines' addresses. Then they in turn pick 'em off the >stack, and so on. Crazy? > >Francis Sounds like a deep blue parallellizing approach :) Well Francis, i fear it's a bit off reality. Deep Blue had a similar approach: first 4 ply: 1 SP processor ply 5..8 : 30 SP processors ply 8..12 : 480 hardware processors above is a similar idea, also not working that well. There are great dependancies: after first move has been searched one can efficiently give other processors a job. Till then they're doing nothing (assuming game start). example problem in your approach: if a processor finds somewhere that it wins material, then isn't it a shame in your approach that they all have done work for nothing, as they all are searching the same gamespace!
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.