Author: Tony Werten
Date: 08:27:36 02/01/01
Go up one level in this thread
On February 01, 2001 at 08:20:39, Severi Salminen wrote: >>>How do you calculate EBF in your program? What kind of EBF you get in Vincent's >>>position (position after 1.e4 e5 2.d4 d5)? I get 3.6 in 9 ply search. I'd like >>>to just compare. >> >>I've found that the easiest way is to see how much time it costs to finish the >>iteration. >> >>Times for XiniX on AMD 333( yes, I know I need a faster computer) >>ply: 5 time: 2s >>ply: 6 time: 3s >>ply: 7 time: 7s ( 2.3 * ply 6 ) >>ply: 8 time: 18s ( 2.6 * ply 7 ) >>ply: 9 time: 41s ( 2.3 * ply 8 ) >>ply:10 time: 98s ( 2.4 * ply 9 ) >>ply:11 time:217s ( 2.2 * ply 10) >>ply:12 time:500s ( 2.3 * ply 11) > >What kind of program is Xinix? BitBoard or array based? Are you using SEE and >futility pruning. 0x88 movegenerator, MVV/LVA move ordering, no pruning except nullmove, normal extensions, programmed in Delphi. >I'm doing a ply 9 search on Celeron300 in 23 seconds so I'd >like to know where does the difference come from? Getting a low branchingfactor takes time. ( A lot ) It's only usefull if you search deep. Your 23s for 9 ply would suggest 82s for 10, 298s for 11 and 1073s for ply 12. Twice as much as XiniX. Of course at short time controls this advantage doesn't come out. ( So I don't play them ) My extensions don't really help to get me to 9 ply fast either. >How many nodes are you >counting each ply? And what about hashtables - are you using them? I use 4 different hashtables. When they are empty my search is very slow: ply speed 5 15 Kn/s 6 31 Kn/s 7 38 Kn/s 8 46 Kn/s 9 49 Kn/s 10 51 Kn/s >Of course >evaluation has a big impact here too. Yes, my evaluation takes more than 80% of executiontime. Tony
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.