[gecode-users] independent subproblems

[Guido Tack]

Hi everyone,

Decomposition during search (DDS) is now part of the Gecode trunk.  If  
you want to try it out, check out the trunk and configure with the -- 
with-boost-include switch so that the boost graph library is found.

To use DDS in your own models, just replace the call to branch with a  
call to decomposingBranch (with the same arguments).  Then you can use  
Gist as the search engine, which will display decompositions as  
specially marked nodes.  Or you use the countDDS engine provided in  
the search module.

Current limitations (things I haven't ported yet):

  * Global constraints (like distinct) are not decomposed
  * The decomposing branching must come last and cover all variables  
not covered by any earlier branching
  * No search engine that enumerates full solutions yet
  * Only IntVars and BoolVars
  * Probably still buggy ;-)

I'm interested in your feedback!

Cheers,
	Guido

Malcolm Ryan wrote:

> Which version does it work with? I was planning on changing to the C++
> version anyway. The only feature of 2.1.0 that I use is the
> extensional() constraint.
>
> Malcolm
>
> On 10/04/2008, at 3:24 PM, Guido Tack wrote:
>> Malcolm Ryan wrote:
>>
>>> On 09/04/2008, at 11:12 PM, Guido Tack wrote:
>>>> Malcolm Ryan wrote:
>>>>
>>>>> Reading your paper that sounds like exactly what I'm looking for.
>>>>> It's
>>>>> still at the experimental stage, I take it? Any plans to integrate
>>>>> it?
>>>>
>>>> Plans, yes, time to do it, no ;-)  No, seriously, we do have plans
>>>> to at least ship it as a contribution that you could download and
>>>> try out, but I can't make any promises when that's going to happen.
>>>
>>> So the line in the paper saying "The source code ... is available
>>> from
>>> the authors upon request."?
>>
>> That's of course no problem!  But it's not integrated into the
>> latest Gecode, that's how I understood your question.  It uses an
>> older version and is not as generic as we could do it now.  And it
>> does not work with Gecode/J.
>>
>> Cheers,
>> 	Guido
>>
>
>
> _______________________________________________
> Gecode users mailing list
> users at gecode.org
> https://www.gecode.org/mailman/listinfo/gecode-users