Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: HW based Crafty

Author: Slater Wold

Date: 18:24:58 03/30/02

Go up one level in this thread


On March 30, 2002 at 19:16:12, Keith Evans wrote:

>On March 30, 2002 at 15:18:13, Slater Wold wrote:
>
>>On March 30, 2002 at 08:46:40, Sune Fischer wrote:
>>
>>>On March 30, 2002 at 03:07:29, Slater Wold wrote:
>>>
>>>That sounds real exciting :)
>>
>>I hope so.  ;)
>>
>>>But do you think you can reach 10M nps by just hardware coding the movegen?
>>>I think the eval() is what's holding Crafty from reaching those numbers.
>>>I haven't profiled Crafty, but I wouldn't be surprised if the movegen is perhaps
>>>only 10%.
>>
>>Every engine I have profiled move gen and move ordering account for 1/2 the
>>program.  If we take away that half, we would double the speed of crafty.  So
>>1.8M nps is now 3.6M nps.  10M was a guess, and a sloppy one at that.  :)
>>
>
>One question I have, that Hyatt could probably answer off the top of his head,
>is what effect your move ordering will have on the effective branching factor of
>Crafty. I assume that you'll be using a MVV/LVA approach? Which is different
>that what Crafty does. This is something that needs to be considered.

The effecting branching factor of Crafty is about 2 or 3 in standard middlegame
positions.  (DB's was 6.)

MVV/LVA was used in every HW based chess program I can think of.  This is
something I have spoken with several people about, and not as of yet figured
out.

>Maybe I'm misunderstanding what you are actually going to do?
>
>Regards,
>Keith



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.