Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: MTD is a big win.

Author: Vincent Diepeveen

Date: 17:42:32 07/20/99

Go up one level in this thread


On July 20, 1999 at 17:31:44, Dan Homan wrote:

>On July 20, 1999 at 15:19:52, Don Dailey wrote:
>
>>On July 20, 1999 at 00:45:58, hgkjhg wrote:
>>
>>>Don, is Cilkchess going to be released as a commersial or freeware program some
>>>time?
>>
>>I thought about releasing Occam when it gets good.  But I don't want
>>to see 10000 occams on the chess server or at tournaments!   I may release
>>a version before it gets too strong perhaps just an executable for linux
>>and windows.
>>
>>Probably there will be a parallel version of Occam that will be released
>>with the Cilk distribution as an example with source code.  Cilk is free
>>and so would this be.  But I don't know if I would release a really
>>advanced version.  You will be surprised how simple it is to write a
>>parallel chess program when you see this.
>>
>>- Don
>
>Don,
>
>I was looking over the Cilk Pousse web page and it was pretty clear from
>this how to use Cilk to write a parallel chess program with minimal
>effort, but I had a question:  Do you need to have local data structures
>(position, move list) at each node?  It seemed to me (as I read the
>web page) that you would.  My assumption is that this would create
>substantial overhead that would be a problem for non-parallel versions
>of the program.

Excellent question. My program is not recursive,
but completely iterative (that's better for parallellism,
see ICCAJ march 1997 article of bob about DTS).

How do i use that in combination with cilk?

>Second question: Can cilk be compiled and used with C++ or does it require
>pure ANSI C?
>
> - Dan



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.