Author: leonid
Date: 04:00:11 07/02/01
Go up one level in this thread
On July 02, 2001 at 03:17:49, Paul Byrne wrote: >On July 01, 2001 at 00:25:52, Angrim wrote: > >>On June 30, 2001 at 06:47:19, leonid wrote: >> >>>Hi! >>> >>>If you are in mood to solve some mate position, look into this: >>> >>>[D]Q2nkn1q/1Q1rr1q1/2QQqq2/b2Qq2b/2NqQN2/2q2Q2/1q1BB1Q1/q2RKR1Q w - - >>> >>>Please indicate your result. >>> >>>Try to indicate all basic parameters about your hardware and name your program. >>>If your program have some Web Site, please say its address. If something new was >>>done in order to speed your mate solving brain, feel you free to describe your >>>improvement. At least one person will read it with big interest for sure. >>> >>>Thanks, >>>Leonid. >> >>Hey, looks like I got the first reply this time :) >>hardware: Athlon 650mhz >>heuristic: PN^2 with transpositions >> >>proved that move f3xh5 wins, 12 turns >>PN2:17137399 evals, 368084 expands, 148.10 seconds >> >>Angrim > >G2K's PN search says: >: i=203538 n=6559485 p=0 d=1000000000 t=97.170 >: Qfxh5 wins! > >Which, if you do the math, shows how horribly inefficient G2K is :) >About 67K nps on a 1.2 GHz athlon... It could be that 67k indicate every efficent search. Less is your NPS for brute force, more efficent is your search. At least, it is something that is true for my program. It could be otherwise for other programs. Sometime for positions similar to the last one, I see in brute force search only around 35k per second. This position is 9 moves deep. For 8 moves and by brute force mine went 93k per second. >G2K's PN2 search takes 198.28 seconds and 13965004 nodes to see it. Time look like excellent for this position. I doubt that any professional program will do any better. I will go and look (just by curiosity) how much time it will take for some programs to solve it. At least, number of moves here is very moderate, it not even reach 100. So, "usual" program will easily accept position to look into it. Leonid. >-paul
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.