Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Chess over LAN revisited - APHID

Author: Vincent Diepeveen

Date: 08:43:14 09/17/01

Go up one level in this thread


On September 17, 2001 at 11:18:43, Robert Hyatt wrote:

>On September 17, 2001 at 06:47:39, Vincent Diepeveen wrote:
>
>>On September 17, 2001 at 05:07:55, Gian-Carlo Pascutto wrote:
>>
>>The problems are huge. I do not know who wrote the 6-7
>>crap. Crafty AFAIK never was running over a network properly.
>>
>>Bob can answer this better of course, but it's complete nonsense
>>that you get 6.0 - 7.0 out of 16 processors at a 100mbit network.
>
>I don't agree.  They _did_ it.  Schaeffer _did_ it.  Others have done
>it also...

If in worldwar I the biplanes were THE way to do dogfight, it doesn't
mean that for nowadays supersonic fighters (i would nearly write down scram
fighters) the way to go is biplane.

That's a very good compare with the experiments from Schaeffer with
todays computerchess techniques used.

>
>>
>>Most likely the speedup is for a crafty version with more extensions
>>WITH shared memory. speedup is in general worse when the searchspace
>>you search is not overlapping for each processor.
>>
>>>Hi all,
>>>
>>>I missed most of the relevant thread but after quickly
>>>scanning through some posts I wonder why nobody has
>>>brought up APHID.
>>>
>>>The reported speedups for Crafty (version 12.something)
>>>with 16 processors are 6-7 without shared memory, and 9
>>>with shared memory.
>>>
>>>Aditionally. the library itself as well as a 'distributed'
>>>Crafty that uses it are already available *now*. It shouldn't
>>>be too hard to adapt the very latest Crafties to it either.
>>>
>>>Whats the general opnion about it? Nobody experimented
>>>with this Crafty?
>>>
>>>--
>>>GCP



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.