Author: Timo Saari
Date: 05:56:36 10/05/01
Go up one level in this thread
On October 04, 2001 at 19:40:20, leonid wrote: >On October 04, 2001 at 14:01:17, Timo Saari wrote: > >>On October 04, 2001 at 10:59:24, leonid wrote: >> >>>On October 04, 2001 at 01:52:13, Timo Saari wrote: >>> >>>>I am rewriting an old chess problem solver program and as of now it can solve >>>>normal mate in n moves type problems when n is not too large. It does a brute >>>>force type search and uses a hash table and killer heuristic. I assume that in >>>>order to make it reasonably fast with more challenging problems (n>6) I need to >>>>implement at least move ordering and perhaps special handling of plies near the >>>>leaves of the search tree. >>> >>>Can you explain what was your chess problem program? When it was written and >>>how? What language you use and what are your plans about your next program as >>>hardware and system expectation? >>> >> >>The original program was written in late 70s in Algol language for Burroughs >>B6700 mainframe. It did solve problems but was not very efficient. Some time ago > >This is very interesting! I have impression that I know absolutly nothing about >Algol language or B6700 mainframe. What were those mainframes at that time? 64 >bits machine? And at what speed? Mhz? Do we have some of those computers still >at work? The B6700 was an ancestor of the still alive NX/LX-series of computers by Unisys. It is based on stack architecture with memory of 48 bit words (+ 3 or 4 tag bits) and it never was a number cruncher but more suitable for business like applications. Algol(in some variations) was/is used as system programming language; there is no assembler level language. Timo > >Cheers, >Leonid. > > > >>I decided to start to learn to program in c/c++ especially for Windows >>environment and when thinking of a suitable project I remembered the old >>program. I have written a GUI (a visible chessboard and basic position handling) >>and adapted the old program to be the search engine. When solving the problem >>the program produces an expandable search tree (limited depth) that can be used >>to show the position corresponding various tree nodes. >> >>My plan now is to try to make the search faster for more challenging problems. >> >>>I think that I can say about mate solver something that I saw before I started >>>writing it. Doing it was nothing more but putting into work something that look >>>obvious. Will say you anyway, maybe it will be useful. >>> >>>1) Checking moves goes first. >>> >>>2) Specilized writing of two plys. >>> >>>When you have to find mate just two plys deep: >>> >>>a) attacking side produce only checking moves. >>>b) responding side should find only one legal move to stop search. >>> >> >>Your points do indeed confirm also my initial plans; actually points a) and b) >>are in the program already and perhaps the first new thing to implement should >>be 1)(as adviced by Heiner Marxen as well). >> >>Thanks, >> >>Timo >> >>>Expect very soon see your program at work! >>> >>>Cheers, >>>Leonid Liberman. >>> >>> >>>>Can somebody point me to any existing information on abovementiond topics? >>>> >>>>Any help appreciated. >>>> >>>>Timo
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.