Moreover, we are the copyright holders, and can do whatever we like with the source code. The GPL license applies to users of the software, not to the original authors. We could even make the next version closed source, if we wanted (but of course, we don't).mcostalba wrote:If we _distribute_ or _release_ the derived tunable version of SF we have to release with sources. But GPL does not forbids anybody to modify and keep for it's internal use the modified code.
A paper about parameter tuning
Moderators: hgm, Rebel, chrisw
-
- Posts: 1808
- Joined: Wed Mar 08, 2006 9:19 pm
- Location: Oslo, Norway
Re: A paper about parameter tuning
-
- Posts: 1822
- Joined: Thu Mar 09, 2006 11:54 pm
- Location: The Netherlands
Re: A paper about parameter tuning
Without saying i agree with anything that Marco Costalba / Pradu Kannan wrote, why do you release a version of stockfish source code that obviously is completely different than the versions you have on the cluster to tune?Tord Romstad wrote:Moreover, we are the copyright holders, and can do whatever we like with the source code. The GPL license applies to users of the software, not to the original authors. We could even make the next version closed source, if we wanted (but of course, we don't).mcostalba wrote:If we _distribute_ or _release_ the derived tunable version of SF we have to release with sources. But GPL does not forbids anybody to modify and keep for it's internal use the modified code.
As Marco/Pradu will have to do a lot of effort to maintain a bunch of code branches.
Vincent
-
- Posts: 292
- Joined: Tue Jul 07, 2009 4:56 am
Re: A paper about parameter tuning
Didn't Marco just explain this? They're maintaining separate versions to avoid disclosing their tuning methodology.diep wrote:Without saying i agree with anything that Marco Costalba / Pradu Kannan wrote, why do you release a version of stockfish source code that obviously is completely different than the versions you have on the cluster to tune?Tord Romstad wrote:Moreover, we are the copyright holders, and can do whatever we like with the source code. The GPL license applies to users of the software, not to the original authors. We could even make the next version closed source, if we wanted (but of course, we don't).mcostalba wrote:If we _distribute_ or _release_ the derived tunable version of SF we have to release with sources. But GPL does not forbids anybody to modify and keep for it's internal use the modified code.
As Marco/Pradu will have to do a lot of effort to maintain a bunch of code branches.
Vincent
-
- Posts: 1822
- Joined: Thu Mar 09, 2006 11:54 pm
- Location: The Netherlands
Re: A paper about parameter tuning
Tord obviously is not a psychopath, i met him in real life. He's a nice normal person who is old fashioned in behaviour and modern in relationships. So i asked HIM the question.Aaron Becker wrote:Didn't Marco just explain this? They're maintaining separate versions to avoid disclosing their tuning methodology.diep wrote:Without saying i agree with anything that Marco Costalba / Pradu Kannan wrote, why do you release a version of stockfish source code that obviously is completely different than the versions you have on the cluster to tune?Tord Romstad wrote:Moreover, we are the copyright holders, and can do whatever we like with the source code. The GPL license applies to users of the software, not to the original authors. We could even make the next version closed source, if we wanted (but of course, we don't).mcostalba wrote:If we _distribute_ or _release_ the derived tunable version of SF we have to release with sources. But GPL does not forbids anybody to modify and keep for it's internal use the modified code.
As Marco/Pradu will have to do a lot of effort to maintain a bunch of code branches.
Vincent
Am i making myself clear?
Thanks,
Vincent
-
- Posts: 2684
- Joined: Sat Jun 14, 2008 9:17 pm
Re: A paper about parameter tuning
Marco/Pradu uses git merge ( http://www.kernel.org/pub/software/scm/ ... merge.html ) to almost effortless keep in sync the two development branches.diep wrote: As Marco/Pradu will have to do a lot of effort to maintain a bunch of code branches.
With this info that Marco/Pradu gave to you for free, Marco/Pradu expects you buy him a beer
-
- Posts: 1822
- Joined: Thu Mar 09, 2006 11:54 pm
- Location: The Netherlands
Re: A paper about parameter tuning
We use SVN.mcostalba wrote:Marco/Pradu uses git merge ( http://www.kernel.org/pub/software/scm/ ... merge.html ) to almost effortless keep in sync the two development branches.diep wrote: As Marco/Pradu will have to do a lot of effort to maintain a bunch of code branches.
With this info that Marco/Pradu gave to you for free, Marco/Pradu expects you buy him a beer
-
- Posts: 2684
- Joined: Sat Jun 14, 2008 9:17 pm
Re: A paper about parameter tuning
git is better, especially for merges and code shuffles (pick stuff from one branch and commit in another).diep wrote:We use SVN.
-
- Posts: 292
- Joined: Tue Jul 07, 2009 4:56 am
Re: A paper about parameter tuning
I have to agree, git is pretty great. SVN is ok though; certainly a big step up from CVS.mcostalba wrote:git is better, especially for merges and code shuffles (pick stuff from one branch and commit in another).diep wrote:We use SVN.
-
- Posts: 1822
- Joined: Thu Mar 09, 2006 11:54 pm
- Location: The Netherlands
Re: A paper about parameter tuning
I have GIT of course installed on my windows box, it looks very amateuristic (sorry Linus) under windows. Really total amateuristic.Aaron Becker wrote:I have to agree, git is pretty great. SVN is ok though; certainly a big step up from CVS.mcostalba wrote:git is better, especially for merges and code shuffles (pick stuff from one branch and commit in another).diep wrote:We use SVN.
SVN is the choice of the GUI project team, i'm not capable of any comments with respect to SVN. On the command line i'm not too impressed by it to be honest.
Thanks,
Vincent
-
- Posts: 12540
- Joined: Wed Mar 08, 2006 8:57 pm
- Location: Redmond, WA USA
Re: A paper about parameter tuning
It seems to me that you cannot make any progress at all, unless you have both people with new ideas and people who test things.diep wrote:Sorry for the selective snipping...
Do i read this correct you say that persons who test things are more important than persons who have ideas?mcostalba wrote: In chess engines testing numbers are more important then ideas (because they require much longer times to get properly), and much more important then struggle to be the first. In your paper experimental data is small, not clearly documented and too focused on the attempted direction.
Please confirm if so.
Vincent
Of course, it is possible for one person to do both tasks admirably well, given the proper resources.