Author: Sune Fischer
Date: 12:47:49 05/25/02
Go up one level in this thread
On May 25, 2002 at 12:58:58, Russell Reagan wrote: >On May 25, 2002 at 09:49:50, Gian-Carlo Pascutto wrote: > >>On May 25, 2002 at 09:32:58, Russell Reagan wrote: >> >>>The problem is that you >>>have pseudo-attacks. >> >>No >> >>>Again, this seems like you could calculate pseudo-mobility, >> >>No >> >>You pick the right bitboard depending on the occupation of >>a rank/file/diagonal. >> >>You need to calculate bitboards in advance for every possible >>occupation of rank/file/diagonal. >> >>-- >>GCP > >I realize that you have to calculate the occupation of the rank/file/diagonal, >but calculating the occupation of the rank/file/diagonal involves ray tracing >just like it would in a non-bitboard approach. > >So for non-bitboard approach, the steps would be: > >1. ray trace > >And for bitboards, the steps would be: > >1. ray trace >2. array lookup >3. AND operation > >So I still do not see how this could possibly be faster. Even after this, you >can't calculate absolute attacks or mobility, unless there is some method >someone would like to describe. > >So, I still don't get it obviously. I think you are right, if you only consider this problem. This part of move generation is probably faster in pure 0x88 notation. But the advantage to bitboards is in the ease of finding attacked squares. When doing a O-O castle move for white you need to check that squares E1, F1 and G1 are not attacked by enemy pieces, and that F1,G1 are empty. With bitboards you can do all this with just a few AND-operations, think about how else you would do this and how complicated and slow it would be! This is where bitboard catches up to the 0x88 notation in speed (IMO). If you also backup the attackboards, you can use them again in the evaluation. I don't know much about Vincent's method, I think maybe Uri uses it too. It seems that method will generate more information at every node, whether this is just a waste of time or if you can actually make use of all this information I don't know, depends also on the search and evaluation parts of the program :) -S. >Russell
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.