Author: Robert Hyatt
Date: 20:31:21 02/11/00
Go up one level in this thread
On February 11, 2000 at 19:12:33, Vincent Vega wrote: >I believe that everybody working on chess in a college in the US is missing out >on a very interesting research opportunity. Commercial chess programmers are >also missing out but it may much harder for them. The idea is simple - >distributed chess program. this isn't new. We had ostrich, phoenix, then zugzwang, p.conners, waycool, etc. Distributed search is not trivial. But it can work. I'm designing one for crafty now to use our 8 X quad xeon cluster. It will take some time as portability has to be factored in, but there will be a distributed crafty. You aren't going to use the internet (maybe internet2 will work) to much distributed search as the latency is impossible. But 'clusters' will work. > >The programming effort it would take to convert any chess algorithm into a >distributed algorithm would be considerable, but it would less for a chess >program that already is SMP. The US universities all have very good Internet >backbone connections and a researcher could probably get an easy access to a >server that could handle the traffic. > >I envision something similar to distributed.net (http://www.distributed.net/) >and SETI@home (http://setiathome.ssl.berkeley.edu/) but users would have a >limited period of time to finish processing their data so that the data could be >collected and the move could be made. To prevent cheating by falsification of >data, test units could be sent and other units could be sent multiple times for >validation. > >Only a small fraction of participants in distributed.net and SETI@home would be >required for creation of an extremely strong system that could challenge the >best GMs. The participants shouldn?t be hard to get with an announcement at >Slashdot, etc. Getting a good GM to play at very long time controls (let?s say >one move per day) could be hard but some would probably welcome the publicity, >even if only one sponsor was involved. If the interest generated by Deep Blue >and Kasparov vs. World is any indication, it shouldn?t be hard to get at least >one sponsor. > >Bob Hyatt with Crafty is in a good position to make something like this happen, >as Crafty is one of the very few good research programs, it?s parallel to some >degree, and people consider it strong (very important because we don?t want >users giving up, because they think that they are wasting time using distributed >Sucks client while they could be using Fritz or Rebel and get done in a fraction >of time). However, good organizational skills are the main criteria here.
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.