Author: Tom Kerrigan
Date: 13:06:09 03/29/03
Go up one level in this thread
On March 29, 2003 at 04:08:09, Jonas Cohonas wrote: >Great experiment! > >Could you try the same test but with a different test suite like LCT II and >increase time per move? i think for this to be a valid experiment you should >have some positions in there that the program only solves after a lager amount >of time spent, to see if bigger H tables will help solve time in those cases., >since a large H table should only be really helpful in games with long time >control. > >Regards >Jonas I don't see how solving a position in a long amount of time with a big hash table is different from solving a position in a short amount of time with a little hash table. I ran the positions I had for 2 minutes and got 31% overwrite with a 16MB hash table... increasing the size to 32MB reduced the overwrites back to 27%, so my guess is you need a hash table with size proportional to the number of nodes you want to search... maybe 128MB for a fast computer at standard time controls? Of course, that's assuming the percentage of overwrites is what matters. -Tom
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.