[Scip] FibreSCIP

[Gerald Gamrath]

Hi Miro,

did you implement the copy callbacks in all your constraint handlers?
Without this, those constraints will be missing in the local subproblems
copied to the solvers and this might lead to wrong infeasibilities if
the optimal solution computed in a local solver is not feasible in the
global problem.

There is also a way to do a separate project for your extensions; in
priciple, this would need to be similar to the examples in SCIP, but
extending ug-SCIP instead of SCIP. Unfotunately, there is no example for
this in the release yet, but I'm sure Yuji can send you one and explain
to you what exactly you have to do.

He is out of office this week, but will be back next week, so I hope you
can wait until then.

Best,
Gerald

On 16.11.2013 23:51, Miro Marinov wrote:
>
> Dear SCIPpers,
>
>
> I have extend SCIP with a couple of constraint handlers. When I use
> this extended SCIP to solve a problem in the appropriate format it all
> works well. However, when I try using the UG framework with FibreSCIP
> (linked to the extended SCIP), it returns 'infeasible' with dual bound
> gap 0%. I have no idea what to do. Alternatively, I can do a project
> of my own (pretty similar to the LOP example included in SCIP) instead
> of extending SCIP with the constraint handlers (which will be more
> elegant solution for me) but I have no idea how to take advantage of
> the parallelism provided by UG in that case (In other words how would
> you run the LOP example in parallel via FibreSCIP)
>
>
> Thank you very much in advance and I am looking forward to hear from
> you. (:
>
>
> Kind Regards,
>
> Miro
>
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/scip/attachments/20131121/6384db01/attachment.html>