Author: Michel Langeveld
Date: 10:15:59 11/26/01
Go up one level in this thread
On November 26, 2001 at 06:13:16, Johan Melin wrote: >On November 25, 2001 at 14:04:32, Michel Langeveld wrote: > >>I have profiled my chessprogram and got the following output [see below] >> >>What I don't understand is: >>- filebuf::sync(void) (filebuf.obj) has highest Func+Child %. Is this normal? > >I do not think this is normal. 98%-99% should be in search. >(Unless the profile is on a very small search) > >>- Both alphaBetaWhite + alphaBetaBlack are > 100% > >This is ok, since most time is spent as child of both alphaBetaWhite and >alphaBetaBlack. > >>- % of generateBlackMoveList(struct > % generateMovesForBlackPiece >> > >Are you using generateMovesForBlackPiece anywhere else? >Do you have several uses of generateBlackMoveList, some of which may be inlined? > >/Johan Melin Thanks for you answer. Your reply about too short time did turn on a light. I modified my program to do a 3/4 ply search because else it would take very long. But in that test it seems that the logfile i/o took so much time that it a considerable amount in my search. I'm thinking about to make a special #bench define which is off during profiling. I use my attack function for the following reason: - begin of each node to see if I myself is in check by last opponent move (used for extensions etc) - addMoves to move list: -normal moves (do move, myself in check, ignore move) -promotions (do move, myself in check, ignore move) -castling (do move, myself in check, ignore move) regards, Michel
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.