Author: Robert Hyatt
Date: 10:55:14 01/30/06
Go up one level in this thread
On January 30, 2006 at 08:27:22, Uri Blass wrote: >[D]8/8/pppppppK/NBBRQNRp/nbbrqnrP/PPPPPPPk/8/8 w - - > >Engine: Rybka 1.01 Beta 13b 32-bit (512 MB) >by Vasik Rajlich > 3.00 2:28 +2.46 1.dxe4 (10.542.472) > 4.00 3:11 +2.46 1.dxe4 (13.595.552) > >This is under the shredder gui. > >When I try rybka under the fritz gui and do not wait for analysis and >quit Fritz after few seconds with no result then rybka continue to work after >quitting fritz based on task manafer and the only way to stop her is to wait or >to stop the process of rybka. > >I think that it is not nice behaviour of rybka. >I think that a program should quit when it get the quit command during analysis >in a short time. > >Waiting of 0.1 seconds on fast hardware is ok but not waiting of more time. > >My guess is that it takes rybka long time to respond to the quit command. >If I quit rybka under shredder then rybka does not continue to work and I guess >that shredder simply does not wait for rybka to stop but simply kill her. > >Uri I would bet this is the problem: He might use the same sort of ply-1 move ordering that I use in Crafty, namely doing a pure q-search on each position resulting from making a ply 1 move, then sorting the moves based on that score. The problem is that this position has a truly huge q-search, because at the root ordering, you need to use -inf, +inf as the alpha/beta bounds for that q-search, and that means all captures are "ok". In crafty, I don't start checking for "input" (quit in your example) until the search (real search) actually starts, which means _after_ the ply-1 move list has been ordered. Since this takes a minute or two her, it could be a minute or two before Rybka sees the "quit" command...
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.