Author: Matt
Date: 19:31:51 03/05/99
Go up one level in this thread
On March 05, 1999 at 21:09:11, Bruce Moreland wrote:
>
>On March 05, 1999 at 18:42:51, Matt wrote:
>
>>I am familiar with the basic concept of the 0x88 board representation, but I
>>have not been able to find a more in depth study of it anywhere. I was
>>wondering if anybody could direct me towards a good paper on it, or perhaps some
>>C source code of a chess engine using 0x88?
>>Thank you
>>Matt
>
>I don't think it's been written up anywhere official.
>
>The idea is that you make a 128-element board. a1 is 0, b1 is 1, ..., h1 is 7,
>then there are 8 un-used squares, and then a2 is 16, etc.
>
>The reason you waste the space is so that there will be a spare bit in the
>middle of the board index.
>
>You can regard an index into a 64-bit board as a pair of 3-bit indexes stuck
>next to each other, the low-order three bits are the file, and the high-order
>three bits are the rank. With a 128-element board, you end up with a 3-bit
>file, a blank bit, then a 3-bit rank.
>
>If you have an index, and are adding to or subtracting from it, in order to get
>to a square that is related in the way that a chess piece moves, this blank bit
>will serve as a carry or borrow bit if you go off the left or right side of the
>board.
>
>For instance, you can loop through the moves of any piece as follows:
>
> for (;; index += delta) {
> if (index & 0x88)
> break;
> ....
> }
>
>The 0x80 part will catch cases where you go off the top or bottom of the board,
>and the 0x08 part will catch the cases where you go off the left or right side.
>
>This keeps you from having to have a row of sentinels around the edges of your
>board, and also keeps you from having to keep distinct rank and file values, or
>other nastiness like that.
>
>It ends up being pretty easy to code a move generator using this scheme.
This much I already understand :)
>Another advantage of the 128-element board is that you can subtract the indexes
>of any two squares in order to exactly determine the relationship between them.
>
>In a 64-element board, if you subtract d1 from e1, you get 1, but if you
>subtract h1 from a2 you also get 1, even though the two squares don't have the
>same relationship to each other.
>
>In a 128-element board, if you subtract d1 from e1 you'll still get 1, but if
>you subtract h1 from a2 you'll get 9, which makes sense because a2 is not one
>square to the right of h1.
I understand this as well, but what is the value of 9? i understand h1 is 7 and
a2 is 16, and subtracting them gets the value of 9, but I don't understand how
this will be useful. 9 is also the difference between two squares diaganol from
each other. more below.
>This holds true of any relationship between any two squares. You can make use
>of this to determine if it's possible to get from one square to another square,
>in order to build an in-check routine or a static exchange evaluator.
I am not sure how you would use this to see if its possible to get from one
square to another. ie. if you are calculating moves for a bishop, one of the
offsets is -9... if you hit a2 -9, you hit h1, so how does the bounds checking
work in this case?
>bruce
Matt
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.