Author: Uri Blass
Date: 17:26:15 07/07/01
Go up one level in this thread
On July 07, 2001 at 16:54:55, Tony Werten wrote: >On July 06, 2001 at 17:49:28, Uri Blass wrote: > >>On July 06, 2001 at 15:04:14, John Wentworth wrote: >> >>>To all the chess developers, how many hours of coding did it take you to write >>>your first program from the ground up? Did you develop your own book? >> >>No,I did not develop my own book and not a GUI or a chess program but only a >>move generator. >>> >>>I am currently working on one now and have only about 9 - 12 hours of coding >>>time spent. I am at the point where I can play back and forth with the program >>>until it runs out of book (modified Crafty book). I also spent time developing a >>>nice looking GUI that displays the moves made. Just trying to gage the total >>>time it will take me to complete it. Thanks for your input. >> >>I did not count time but my estimate is total time of 500 hours and I only have >>a move generator and not a chess program and even the move generator is not >>perfect and there are some rare cases when I know that it generates illegal move >>or does not generate legal moves(It never generates en passant captures in reply >>to check and it generates illegal en passant captures when there is no check and >>the pawn is not pinned) but my priorities is first to make it faster and only >>later to correct this problem. > >Based on experience, I STRONGLY suggest to do it the different way around. First >correct the problem then make it faster. >Even more, fixing problems can make your engine faster. Correct movegeneration >might give you more (correct) cutoffs and slim down your tree. You are right when you have an engine that is playing or if you have problems that are because of bugs and not because of something that you did not work about. The point is that some ideas that can make my move generator faster involves changing the structure of arrays or changing the structure of functions and fixing the problems in my move generator now means doing later more changes in my code when I change arrays or change using of functions. Another point is that I have no problem to use the known position that was posted for debugging my move generator because the known mistakes in my move generator do not cause mistakes in calculating perft 5 in the relevant position. I did not think of doing it faster before being able to calculate perft 5 correctly The order that I plan is: 1)making my move generator faster without creating new bugs and I guess that there are not new bugs if it calculates perft 5 correctly in the relevant position. 2)correcting the errors in the move generation. 3)building search and evaluation functions. Uri
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.