Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Perft 5,6 {Fastest program is List}

Author: Sune Fischer

Date: 16:30:53 06/05/02

Go up one level in this thread


On June 05, 2002 at 19:10:40, Dann Corbit wrote:

>On June 05, 2002 at 19:01:42, Dieter Buerssner wrote:
>
>>On June 05, 2002 at 18:55:50, Uri Blass wrote:
>>
>>>It is possible to play the program against itself under winboard without a
>>>special command so I do not see a reason for a special Auto command.
>>
>>An auto command will need less resources. Only once the code will be in memory,
>>and the same for the hash tables (if you assume an engine with assymmetrical
>>evaluation some complications may arise). So you can use double hash size at
>>least. Also, it may be much more easy to debug/profile/... a stand alone session
>>of the engine, than the engine running under some GUI.
>
>Besides, if you have only one CPU, half of the time will be idle waiting.
>That will really skew the time resources and also completely hide the
>bottlenecks (probably reading from stdin will look like the hot point).
>
>Playing against yourself also means that you do not have the same hash table.
>Your program will compute faster if it always guesses exactly right and does not
>have to compute each hash value twice.

You can profile it in analysis mode.
Sending and recieving moves is not going to show up on the profile anyway.
I have three standard positions I use for profiling, one opening, one middelgame
and one endgame position.
These have vastly different profiles.
I like that idea more than profiling an entire game, a game that will probably
change everytime you change something in the code, so you can't compare it to
previous profiles.

-S.



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.