Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Tweaked tscp at my site

Author: Dann Corbit

Date: 14:09:33 12/07/99

Go up one level in this thread


On December 07, 1999 at 08:16:15, blass uri wrote:
>On December 07, 1999 at 02:32:47, Dann Corbit wrote:
>
>>On December 07, 1999 at 02:20:40, Dave Gomboc wrote:
>>>On December 06, 1999 at 23:28:28, Dann Corbit wrote:
>>>>I messed around with the source code, so don't bother Tom with any bugs you
>>>>find. Sorry, I just can't help myself.
>>>
>>>Reordered (e.g. using your ".cpp" includes), or altered?
>>Both.  I changed some data structures from int to unsigned int, changed the
>>ordrer of evaluation in if() to short circuit in a more favorable way.  About
>>20% faster than the original, I think.  Some simple changes will probably make
>>an even bigger impact, including one I am discussing with TK right now about
>>evaluation of check.
>
>I believe that it is possible to do Tscp faster.
I think that is the whole idea.  It's like Don Daily's old Tic-Tac-Toe program
he posted here.  20 lines or so, but it was plenty to get you going and even had
alpha-beta.  If you throw crafty at some new beginner, they're going to blanch.
Even Phalanx is a mouthful.  But TSCP can be read and comprehended in an hour or
so.  It makes chess programming much less intimidating.  I have seen code for
other programs that clearly started out as TSCP at some point and then grew.

>I told Tom Kerrigen that it is possible to save time in the attack function by
>adding matrix to the program  but he told me that the idea of his simple chess
>program was to do a program that is simple to understand and not to do the
>program faster
I have been conversing with him also with ideas.  But he is right.  Instead of
creating a nuclear reactor powered starship, it is intended as a rowboat.  If we
want something complex and faster, there are more complex and faster
alternatives.  The nice thing about it is that we can make a few simple changes
and see how it performs.  It could even be a nice testbed for testing out ideas
before implementing them in a more complicated program.

>I guess that it is possible to do the program 500% faster if it is the target.
>
>The question is what do beginners prefer?
There is no answer to this question.  I prefer simple because I found it easy to
instantly understand.  For advanced stuff, I look at the clever evaluations in
Phalanx and Crafty.  "What is the goal of this tool?" we might ask ourselves.
The answer (in the case of TSCP) is "Show someone how to write a minimal chess
program in a way that is easy to comprehend."  I think it accomplishes that goal
admirably.  If the goal were to show how rotated bitboards, EGTB's and hashing
can improve chess performance -- we already have the answer to that question:
Crafty.

>I hoped to see a program with no complex evaluation or complex search rules but
>but not a stupid program that you can easily do the same thing faster but maybe
>other beginners do not care about speed.
Of course they care about speed.  But before learning how to run, they have to
learn how to crawl and then to walk.




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.