Author: Robert Hyatt
Date: 16:59:57 09/08/04
Go up one level in this thread
On September 07, 2004 at 23:56:26, Stuart Cracraft wrote: >On September 06, 2004 at 22:41:32, Robert Hyatt wrote: > >>On September 06, 2004 at 22:05:58, Stuart Cracraft wrote: >> >>>On September 06, 2004 at 18:33:47, Robert Hyatt wrote: >>> >>>>On September 06, 2004 at 04:47:12, Alessandro Scotti wrote: >>>> >>>>>On September 05, 2004 at 22:12:35, Stuart Cracraft wrote: >>>>> >>>>>>I walk the PV until there is no further position. No limit. This can then >>>>>>either be left in the hash table if you don't clear between searches >>>>>>but like Bob, you should keep the PV entries and restore them in the >>>>>>hash table before the next search whether or not you table-clear due to >>>>>>potential changes. >>>>>> >>>>>>I don't do this. It's a simple fix. I don't know why I've delayed putting >>>>>>it in. >>>>> >>>>>I don't either. But since I get my PV from the hash table, it's already there. >>>> >>>> >>>>It won't _always_ be there . That is the problem. You can overwrite a key >>>>position and lose a move in the PV, and hence the _rest_ of the PV will not be >>>>useful. >>> >>>I understand that. But unless you record the PV with an alternate technique, >>>you can't fill it since source == destination. >> >>I was assuming a non-flawed method of maintaining the PV. :) >> >> >> >>> >>>And I am not sure I want to do the triangular array gubbish. >> >> >>That's _the_ answer to do it right, however. And it is trivial to do. > >The method I saw at Bruce's webpage doesn't look great to me. In Crafty, it adds 3 lines of code to quiesce.c, and another three lines of code to the points in search.c where I return a value other than alpha or beta...
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.