Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: 64-bit machines

Author: Robert Hyatt

Date: 13:05:23 02/10/03

Go up one level in this thread


On February 10, 2003 at 14:46:12, Tom Kerrigan wrote:

>It occurs to me that your test does not take temporal locality into account. Why
>not run your test and clear the table every 100 nodes or so? Then you have a
>better idea of how much is staying in cache.
>
>-Tom


The problem is that for those attack tables, they are used _frequently_.  IE in
move
generation.  In the SEE code.  In the evaluation code.  IE I just ran one simple
test to
check, and it looked at 2M nodes.  The total counters for those "tables" was up
in the
200M range...  That is the sum of all entry counters.  For a single node,
obviously I
need at least 8 entries for the pieces for the side on move.  plus several for
each move
ordered by SEE, plus several for each evaluation, etc...  It grows pretty
quickly.  That
was 200M references to 8 byte values, in about one second.  Obviously part of
that has
to be in cache.  But just as obviously not all as my dual xeon has 512K L2 only.



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.