Author: Tony Werten
Date: 22:29:12 03/01/05
Go up one level in this thread
On March 02, 2005 at 00:22:05, Joshua Haglund wrote: >On a P4 3.0 Ghz. > >All 3 & 4 man take 15 minutes to generate. > >No problem. > >5 man, they could take anywhere from 20-some seconds to 14 hours each. As far as >I've seen myself anyway. > >It it possible to generate 6-man, a part at a time and be able to use that part >even though it is imcomplete. It is certainly possible (been doing it myself) but there is one side effect you have to take into account. Probing the disk is very expensive. It pays itself back because you can skip part of the tree when that tree is big enough. That's why you only probe when there is sufficient depth left. Probing incomplete tables runs the risk that you do all the expensive stuff, and then come back with no result. So you have to mark these incomplete tables and only probe them at a certain remaining depth wich is higher than your normal (into complete tablebases ) remaining depth. Tony > .... >knnknn wtm: Starting major iteration 1 >Generating... > knnknn.a.nbb >knnknn btm: Starting major iteration 1 >Generating... > knnknn.a.nbw >knnknn wtm: Starting major iteration 2 >Generating... > knnknn.b.nbb >knnknn btm: Starting major iteration 2 >Generating... > knnknn.b.nbw >knnknn wtm: Starting major iteration 3 >Generating... > knnknn.c.nbb >knnknn btm: Starting major iteration 3 >Generating... > knnknn.c nbw > .... > knnknn Generation done. > Combining iterations... > copy knnknn.a.nbb+knnknn.b.nbb+knnknn.c.nbb knnknn.nbb > copy knnknn.a.nbw+knnknn.b.nbw+knnknn.c.nbw knnknn.nbw > del knnknn.a.nbw knnknn.a.nbb... > del temps... > >First it would create temp.a.nbb, temp.a.nbw. Once those are done, temp = >knnknn.a.nbb, knnknn.a.nbw. > >I know there is a bit vector mode for continuation of interupted generation, but >allowing the use of parts might be helpful. Atleast I think it would be. :) > >Joshua Haglund >toneewa@yahoo.com
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.