Mikael Zayenz Lagerkvist <zayenz@...> writes: > > Hugh Ferguson skrev 8/21/13 5:16 AM: > > Gecode looks attractive, aside from the lack of a > > Java interface. Can anyone give me an > > idea of whether it would be worth using Gecode/J as a starting point, > > or whether it would be better to start from scratch? > > I agree with Guido's comments on not to reuse Gecode/J. > > However, depending on what your particular use case is, building a > direct Java interface may be way to general. If what you need is to > solve a single optimization problem, then it might make more sense to > build the model in C++ and to interface that model with the rest of your > system. This way, the only thing that is needed to communicate between > Java and C++ is the model instance and the result. This way, it is > possible to run the Gecode part in a separate process that will not > interfere with the JVM. > > Cheers, > Mikael > Mikael
Thanks to you and Guido for the quick response. Unfortunately, I am dealing with literally hundreds of problems, although they are all similar. In some cases I must build and solve 50,000+ model instances in a day, although they are all fairly small problems. I need to be able to build models "on the fly" from a more abstract specification - perhaps FlatZinc. _______________________________________________ Gecode users mailing list users@gecode.org https://www.gecode.org/mailman/listinfo/gecode-users
