Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Tuple-based processing

Author: Dan Andersson

Date: 13:23:08 11/10/04

Go up one level in this thread


On November 10, 2004 at 12:43:55, E. Nielsen wrote:

>On November 10, 2004 at 11:35:39, Dan Andersson wrote:
>
>>... One problem with Tuple spaces is that it introduces a
>>shared memory to any algorithm.
>
>And... Why is this a problem?

 Several reasons :) Here are two.
 A shared memory in a multithreaded environment is a bottleneck. This can be
somewhat alleviated by distributing the shared state. But it's only hiding the
overhead.
 And reasoning about an algorithm becomes a bother when the environment includes
global state. Especially one as inherently serial as alpha-beta. The bigger
state the worse it gets. It's not all that bad if it is possible to isolate a
smaller critical state. But then the idea of Tuple space seems like overkill.
Some other abstraction is probably more appropriate.

MvH Dan Andersson

>
>Regards
>E. Nielsen



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.