Re: [computer-go] Help with gmp

[Don Dailey <drd@xxxxxxx>]

> I am aware of GTP, but I choose to use GMP for two reasons: 1) I already have
> the code that implements GMP, and 2) Ill probably need to play against other
> programs that only implement GMP.  Why do you say I should not use it? Only
> because the existance of a better one or is there some other reason?
> 
> Thanks for your help.


I'll try to answer that even though I didn't say it.

First of all, you'll probably  want to debug things.  That's very easy
with GTP.  It's very hard with GMP.

GMP doesn't  have much support  for anything other than  passing moves
back  and forth,  so you  have to  set up  everything  manually before
playing each game.  There are probably workarounds.

You don't have the code to implement GTP?  You only need a few minutes
to write the code yourself.  It's easier to write this code than it is
to  understand the  interface to  the  GMP library  (which isn't  hard
either with the library.)

You are  worried that there are  other programs that  do not implement
GTP but it will soon (if not already) be the other way around.  My own
program doesn't support GMP but it does support GTP.   

I know  of only one program  that doesn't support GTP  and that's Dave
Fotland's  program because  he stated  so.  There  could be  a million
others, but I don't know about them.

The biggest  reason you shouldn't use  it, (and there are  still a few
diehards who will  probably take offense), is that  it's just crazy to
NOT use something that is far more useful AND far simpler.  If you are
building tools to support automated weight tuning via neural networks,
something  that you  will be  experimenting with  a lot,  you  will be
hurting yourself not to use GTP.

- Don
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://computer-go.org/mailman/listinfo/computer-go