[gecode-users] Extending Integer Variable

[Gustavo Gutierrez]

On Tue, Apr 8, 2008 at 2:52 AM, Filip Konvička <filip.konvicka at logis.cz>
wrote:

> > I am also interested in implementing FloatVars. I already built a solver
> which
> > works on floating variables using the MPFR library (http://www.mpfr.org/).
> I
> > designed some propagators
> > (+,-,*,/,max,min,cos,acos,sin,atan,asin,exp,log,sqr,sqrt) which work
> quite
> > well on my benchmarks.
> > I came to GeCode recently because of all the work done for the integer
> and set
> > variables, but I will be glad to share my code with anyone who is
> interested
> > in implementing FloatVars in GeCode.
>
> MPFR is nice, but unfortunately it is licensed under LGPL, which I'm
> trying to avoid. I think that for my purposes Boost.Intervals operating
> on 'double' should be sufficient.
>
> But let's not waste effort on implementing FloatVar twice! I have
> limited time for all this myself, but with the help of my colleagues I
> think we can try to do some initial sketch implementation, while we'll
> try to make the domain implementation part parametrized, so we can
> switch between Boost.Interval, MPFR, CUDA :-D or whatever.
>


There are some other implementations that may result useful. One is the
interval library that comes with boost (already mentioned), also there is an
implementation by T. Hickey called smath (
http://interval.sourceforge.net/interval/prolog/clip/clip/smath/README.html)
and a complete solver for continous domains called elisa (
http://sourceforge.net/projects/elisa/).

Smath provides a wide set of narrowing operator implementations and claims
to be correct, i have used it before and its quite good. Elisa comes with
different representations for constraints on continous domains (natural
interval form, newton interval form, etc.) and that makes it promising.
Boost.interval is the one that best deal with portability which is a matter
when dealing with floating point numbers.

I think there are very good sources  of information (papers and
implementations) and gecode could benefit of all of them.


>
> Personally, I am all for contributing and sharing code. From our
> company's point of view, however, we must be sure that we can actually
> benefit from that - and basing our work on LGPL code is more risky (and
> uncomfortable) for us. That's why I prefer the MIT-style licenses.
>
> Cheers,
> Filip
>
>
> _______________________________________________
> Gecode users mailing list
> users at gecode.org
> https://www.gecode.org/mailman/listinfo/gecode-users
>



-- 
Gustavo Gutierrez
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.ps.uni-sb.de/pipermail/users/attachments/20080408/d26dfed8/attachment.htm>