Author: Gian-Carlo Pascutto
Date: 11:44:50 09/06/00
Go up one level in this thread
On September 06, 2000 at 07:34:48, leonid wrote: >Hello! > >Please, solve this position and say how much difference your hash table did. I do not use a hash table but store the complete tree in memory. This means I can reuse results from former searches, but I cannot recognize transpositions. >Indicate any parameter of your search that you consider important. My approach is similar to conspiracy numbers. > >It is pretty normal position from the book. > > kb6/B3P3/K1P5/p7/P7/3r4//8/8 w > >My solver could not resolve it by selective search and went paintful 24 sec (AMD >400) before finding the mate. Mine have no hash tables. When I went to see it on >Rebel 10, I was impressed. His time in solving the position knew around 6 fold >improvement between solution without hash and with it. His hash was 28 M. > > q1bkb1q1/2nqn3/1BrnrB2/1R1Q1R2/1NNQNN2/2Q1Q3/2QQQ3/3K4 w Sjeng's matefinder, which uses a best-first, selective algorithm finds the mate in 7.2 seconds, 56682 nodes and 2.8MB of mem used on my Cyrix120. Normal Sjeng finds the mate in 11.6 seconds, 392217 nodes. (16M hash) Turning OFF the hash makes it find the mate in 6.7 seconds and 201233 nodes. Argh! >This last, is the fantasy position. Did it yesterday. Also could not solve it at >minimum number of moves by selective search. It took lengthy 18.5 min to reach >minimum number of moves for mate by brute force. Selective search can reach it >only if one move more is permitted. Then it take it in 2 sec. But how much hash >table could help in this position? 158 seconds on my Cyrix120. (641784 nodes, 31MB of mem used) Normal Sjeng didn't find it before my patience ran out. -- GCP
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.