[gecode-users] Concept of "computation spaces" in Gecode vs ILOG.

[Pieter Thysebaert]

Hello,

As some of you may have learned from a previous posting, I'm currently
busy implementing a parallel constraint optimization framework.
The ultimate goal is to run this framework on a cluster of PC's (i.e. a
distributed memory system).

Using gecode and MPICH2, I've succeeded in creating a working prototype.
Some reasons for this modest succes are 1) the availability and
licensing of Gecode and MPICH2 and 2) the concept of computation spaces,
which makes it easy to encapsulate the constraint store and propagation
things and reuse these concepts in a custom program.

As my employer currently has a ILOG Solver license (sorry ;-) , the
question  arises however if it would be technically feasible (i.e.
disregarding licensing issues) to implement the same framework using
ILOG Solver technology. I have only used ILOG Solver the ILOG way
(modelling variables and constraints, setting goals, custom selectors
and choosers, extracting into a solver object ...).
My question is thus: is anyone here familiar enough with both products
(Gecode and ILOG Solver) to tell me whether the ILOG Solver internals
can be manipulated in the "computation space way", either thorugh
existing ILOG API or some wrapper classes?  Does anyone have experience
with using ILOG Solver in a PC-based distributed memory system (as far
as I  can tell, ILOG Parallel Solver is limited to shared-memory
multiprocessors?)?

Thanks,
Pieter