Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A nice speedup - maybe (?)

Author: Pham Minh Tri

Date: 18:04:05 03/08/01

Go up one level in this thread


On March 08, 2001 at 11:24:48, Robert Hyatt wrote:

>On March 08, 2001 at 10:23:54, Robert Hyatt wrote:
>
>I have now done some testing.  First, a programming issue.  If you call
>Quiesce() note that tree->sort_value is used to sort the captures.  I use
>this same array to sort the root moves.  So as you call Quiesce() you will
>corrupt the values you have already found for the first few root moves.
>The solution is to declare a local sort_value[256] array and change _all_
>references to tree->sort_value to just sort_value.
>
>After you do that you will find that the search is no faster than before,
>as the original sorting scheme was very accurate.  The plus for using
>Quiesce() is that the code is a bit shorter however, but it is also a tiny
>bit slower as the q-search is a bit slower than the way I was sorting the
>moves before.
>
>Basically it is a break-even proposition.  No big savings or loss at all,
>over a 24 problem test comparison...
>
>Bob

I agree that use of Quiesce in this case makes some problems. Actually, I have
just rewritten another function of Quiesce for that purpose - it means no
savings of code compare with yours. However, it is easier for me to manage it.
And when I try it with much more games, the time savings drop to around 5-10%,
up to games, not big as my first result, but still worth to try.



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.