Author: Guillaume MOYA
Date: 09:01:24 02/22/05
Go up one level in this thread
On February 22, 2005 at 11:36:39, Andrew Williams wrote: (a) No, I've got a P4 2.6C, not the high end standard but it works. (b) Sorry, I should have write "history heuristic", I store in a 64x64 (from-to) table, the cut-off count, then I perform a quicksort on the move generated. I changed my king check detection routine, and it takes 1703 ms to find the PV now, how many time it should take? Guillaume >On February 22, 2005 at 10:25:22, Guillaume MOYA wrote: > >Hi Guillaume, > >If it's taking 3 seconds to complete 4-ply search, I suspect that you have one >of two problems: > >(a) A very slow computer :-) >(b) Some error in your move ordering. What do you mean by "memory heuristic"? > >Andrew > > >>Hi, >> >>I write in my spare time, a little chess program, I just have done the move >>generation and I play with alpha beta algorithm. My move generator is coded >>using the rotated bitboard stuff, it works well but I got very bad speed result. >> >>As an example, with this fen : >> >>r4rk1/p2q1ppp/n1pb4/1p1p3Q/3P2NB/P3P2P/1PP3P1/2R2R1K w - - 0 1 >> >>when searching only at depth 4, with the simpliest eval function (just piece >>value count), it takes 3094 ms to compute the PV. Here's how my program proceed >>: >> >>1) Generate all moves >>2 Order move, using memory heuristic >>3) Expand tree, test if the king is in check, remove the node if yes >> >>and I use plain alpha beta with no funny stuff. When profiling my program, main >>time is consummed in the function that's computed attaqued square (called after >>a move is done to test if the king is in check). Is there another method? Is >>this kind of performance right without a hashtable or something else? >> >>Thanks >>Guillaume
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.