Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: To bitboard or not to bitboard?

Author: Russell Reagan

Date: 20:43:39 08/30/03

Go up one level in this thread


On August 30, 2003 at 22:29:56, Sune Fischer wrote:

>Another, IMHO, really interesting aspect of bitboards is the floodfiller
>techniques.
>There is a whole new area yet to explore here, and as far as I know they
>have no equivalent in non-bitboard environments.

Hi Sune,

I agree about the kogge-stone stuff. I think they are very interesting, and
quite clever. They just go to show that even in such a heavily researched area
as computer chess, there are new things waiting to be discovered.

I'd like to see how the simple C routines will run on an Opteron. If Gerd was
able to get the MMX routines to go faster than his rotated bitboard stuff on
32-bit hardware, I imagine that a C version would run even faster on an Opteron.
Also you have the 128-bit XMM registers, which means you could generate multiple
attack bitboards at once.

The C version of the routines were horribly slow for me on 32-bit hardware, but
I admit that I used them a little foolishly and could have called them far less
often than I did.

For 32-bit hardware, I've stuck with rotated bitboards simply because to get the
kogge-stone stuff fast enough you have to do MMX assembly, and that was a little
bit too tedious for my tastes. I wrote a few routines in MMX assembly, and it
did give a significant speedup, but not close to my rotated bitboard speed. I'm
no assembly programmer like Gerd :)



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.