gnugo-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [gnugo-devel] GNU Go Development Idea


From: Steven
Subject: Re: [gnugo-devel] GNU Go Development Idea
Date: Thu, 3 Mar 2005 15:16:48 -0400
User-agent: Internet Messaging Program (IMP) 3.2.2

GNU Go Developers,

Thank you very much for your response!  I am thrilled to have this idea 
considered.  I
hope the code I have attached can be easily executed, and I could have your 
comments. I
have submitted just the files I have modified as part of this project, the 
remaining
files are all unchanged GNU Go 3.6. If you would like changes submitted in 
another way,
just let me know.

You will want to look at genmove_ponder() in genmove.c first. This is called 
from
play_gmp() via gnugo_genmove_ponder(), meaning GNU Go MUST currently be run 
with --mode
gmp (i.e using cgoban) for the function to be called.

I realize there are quite a few shortcomings with this method, but the core 
idea is
implemented here.  If there is any way to improve on this idea at all, I would 
be glad
to hear it.

Quoting address@hidden:
# I can imagine that if the opponent is playing very quickly
# and not making the expected moves this could cause GNU Go
# to play more slowly, and there could also be persistent
# cache issues.

This is definitely one issue, the pondering does become useless if the opponent 
plays
quickly.  The best situation is one in which the opponent (computer or human) 
is taking
a liberal amount of time to consider it's moves.  Time can be wasted now if GNU 
Go is
pondering a move when the opponent has already responded, and the move given 
has not
been pondered.  But whether this time is really significant depends on the time
controls, I suppose.

The other main drawback is that I used the play_gmp.c interface code to make 
the calls
to the pondering method, since this was the most straightforward way I could 
find of
doing it.  However I came to realize later that these interface files are not 
the right
place for such code, at all.  Any ideas are welcome on a better way.

Finally, note that the files have liberal gprintf's to indicate what the 
computer is
doing.  These are cheap debug statements, but hopefully this makes the operation
clearer for the time being.

Thank you again,

Steven
University of New Brunswick

the attached files are: 
\interface\play_gmp.c
\engine\genmove.c interface.c gnugo.h

Attachment: genmove.c
Description: Text Data

Attachment: gnugo.h
Description: Text Data

Attachment: interface.c
Description: Text Data

Attachment: play_gmp.c
Description: Text Data


reply via email to

[Prev in Thread] Current Thread [Next in Thread]