Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Bob, how about P2P distrubuted Crafty

Author: Robert Hyatt

Date: 18:58:18 10/28/00

Go up one level in this thread


On October 28, 2000 at 18:22:46, Ratko V Tomic wrote:

>A Crafty using TCP/IP and peer-to-peer (P2P) distributed algorithms,
>deploying, say, 100,000 or more processors and hundreds of terrabytes
>of table-base space. There has been rising interest in P2P
>of late, including Intel and O'Reilly P2P conferences (cf.
>http://www.oreillynet.com/pub/a/linux/2000/09/22/p2psummit.html
>also http://www.oreillynet.com/p2p/ ). Intel, in particular seems
>interested in distributed processing application of P2P.
>



This is high on my list, as I have said previously.  I got blindsided by the
wedding plans for my daughter's wedding, which took _all_ my spare time
starting in February, and ending last weekend as we cleaned up the remnants
of the ceremony at my house...

I hope to begin to think about a distributed tree search again.  But note that
using thousands of processors over the internet is really not yet feasible,
as the lag/jitter is way too high...

But clusters can work.  And maybe a WAN type implementation can work for
longer time controls.




>Being the top guru on distributed & parallel chess algorithms since
>the Cray Blitz days, if you have time, that could be an interesting
>project, for you and many others well beyond the computer chess world.
>Then there could be a match Kramnik vs Internet (or vs P2P Crafty
>or vs Net Crafty).
>
>The effective algorithms here would likely be quite different than
>anything seen on regular multiprocessors due to drastically different
>ratios of local/global memory access speeds as well as the huge
>difference in the maximum bulk computing power (or storage) for
>the two configurations.



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.