Author: Robert Hyatt
Date: 21:02:43 10/17/02
Go up one level in this thread
On October 17, 2002 at 22:18:07, Jeremiah Penery wrote: >On October 17, 2002 at 20:50:55, Uri Blass wrote: > >>I do not try to defend vincent explanation but to explain how it is possible to >>get 4 plies by normal definition when 5 is the maximal depth of the hardware. >> >>Suppose that the software searches in some line 3 plies and the remaining depth >>after the 3 plies is 5 thanks to extension. >> >>It means that the normal depth of the search was 4 and the maximal depth of the >>hardware was 5. > >I'm not totally sure, but from what I've read it seems like all software >extensions were resolved in software. That is, it never passes _extra_ >remaining depth to the hardware because of extensions in the software. That >would probably mean a lot of extra times where the hardware can't finish its >search fast enough (remember, they limited a single hardware search to 8000 >nodes), and has to pass back to the software for more plies, then it would get >split back to the hardware again with less depth. That's the way I interpret everything he has said. The SP2 has to be tuned just right to keep the chess processors as busy as possible without wasting time waiting on them or wasting time while they are idle waiting on the SP2... The chess processors have to be called when the remaining tree is small enough to finish within the time-out period they chose.
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.