Author: Robert Hyatt
Date: 18:34:20 08/19/05
Go up one level in this thread
On August 19, 2005 at 19:06:59, Brian Richardson wrote: >From time to time, Tinker loses games on time, even though it is very rare, and >sometimes even with an increment. The text below is from Tinker's log file (not >the Winboard debug file). Notice the "long" (nearly one full second) time >between the time, otim and move messages. Earlier in this game, the time >between these messages was very short. Of course, Tinker was pondering during >this time, so the delay is likely due to Tinker not checking for input >frequently enough. However, just wondering if others have seen anything like >this. > >More importantly, when do you count your time as starting: with the actual >move, or the "time" message which comes earlier? > >Thanks > >13:14:10.671 Got2: time 200 >13:14:10.921 Got2: otim 500 >13:14:11.546 Got2: d7d6 When I watch, the two come within less than amillisecond of each other. Which means this isn't really much of an issue as to whether you start counting when you get "time/otim" or the move. In crafty, I look at the target time per move. If it is several seconds at least (you can look at the code that sets nodes_between_time_checks) then I set this value to approximately my NPS, so that I check the time and check for input about once a second. As the target time for a move shrinks, so does the above, so that I always check the time and input buffer frequently enough that I don't overstep the time for this move by very much... Make it "adaptive" and the problem will probably go away...
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.