Author: Pham Minh Tri
Date: 16:31:44 11/21/01
Go up one level in this thread
On November 20, 2001 at 20:06:34, Dann Corbit wrote: >On November 20, 2001 at 19:37:50, Ren Wu wrote: > >>On November 20, 2001 at 19:17:12, Peter McKenzie wrote: >> >>>Its the method used by GNU Chess. >>>You have a table containing a list of moves for each combination of piece type >>>and square. The contents of this table is typically computed on startup. >>> >>>So for example, a knight on A1 can only move to B3 and C2 so you would be able >>>to index into a big table like so: MoveTable[KNIGHT][A1] and you would have >>>access to a list containing B3 and C2. >>> >>>It works for sliders too, although of course you have several lists - one for >>>each direction and you must check the board for blocking pieces. >>> >>>cheers, >>>Peter >> >>Peter is right. >> >>I am using this method to generate moves. It is more sensitive to memory speed. >>My program runs much better on RDRAM based machines. >> >>I think this method is slower than x88, or even slower than the old mailbox >>method in current generation machines, even with rambus memory. > >Since you can do the same sort of move generation with bitboards, what is the >advantage of using Move Tables then? The main advantage is that it is one of fastest methods on the non-bitboard world. I mean everyone should consider this method if he does not use bitboard. > >Is it extremely easy to understand and debug?
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.