Computer Chess Club Archives




Subject: Re: Matt Taylor's magic de Bruijn Constant

Author: Bas Hamstra

Date: 10:17:56 07/13/03

Go up one level in this thread

It is used *extremely* intensive. Therefore I assumed that most of the time the
table sits in cache. But apparently no... Makes you wonder about other simple
lookup's. A lot of 10 cycle penalties, it seems.


On July 13, 2003 at 12:42:37, Andrew Williams wrote:

>On July 13, 2003 at 12:17:19, Russell Reagan wrote:
>>On July 13, 2003 at 07:43:26, Andrew Williams wrote:
>>>Probably a stupid question, but if you have a loop which does just LastOne with
>>>this scheme, wouldn't the table get cached and therefore make it go much faster
>>>than in a test where it's embedded in your code?
>>Why would a 64-byte table not be cached in the real run of the chess program?
>I always assume that it would be pushed out in the course of all the other
>things that the program is doing. I guess it depends on how frequently the table
>is used. I don't know anything about bitboard programs, so I couldn't begin to
>judge what is happening. One thing I can recommend is valgrind and cachegrind:
>These seem to be Linux-based tools, but I assume you can get something similar
>for Windows. Cachegrind tells you how many cache misses every function in your
>program causes. Cachegrind is actually an option to valgrind, so you run
>valgrind with a special option to do the cache checking.

This page took 0.1 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.