Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: movegen speeds(was Re: Status of Brutus?)

Author: Vincent Diepeveen

Date: 14:30:50 07/29/03

Go up one level in this thread


On July 29, 2003 at 17:03:08, Tord Romstad wrote:

>On July 29, 2003 at 17:00:05, Vincent Diepeveen wrote:
>
>>perft goes at 1 billion moves a second when using a hashtable as we discussed
>>here in CCC.... ...if it just runs long enough
>
>This is probably an extremely stupid question, but I still dare to ask it:
>What is this "perft" thing which everybody talks about?
>
>Tord

from openings position they recursively generate all the legal moves or
something and just give back a legal move generator count. nothing stored
therefore.

with a transpositiontable you can keep counts. like how many moves will get
generated in the subtree below it.

so the deeper you search the more this transpositiontable will bring (if it is
big enough especially).

in the long run it will of course give huge boost to it because it improves the
branching factor.

then the last ply you don't need to generate the moves in fact. you just need a
verification whether some moves are legal or not.

then there is other tricks you can try.

all those tricks you cannot use for a normal chessprogram.

what i am talking about is a normal move generation (semi legal) from my normal
move generator.

Note that Yace and DIEP are about equally fast generating moves when we compared
on some old hardware. I guess at P4 and K7 the yace is a lot faster nowadays too
if Dieter has such a machine nowadays.








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.