[Scip] Some questions about the LPI interface

[Ambros Gleixner]

Hi Daniel,

yes, Cplex will automatically warmstart in this case.

If you want to see Cplex output in order to confirm this, you can call

    SCIP_CALL( SCIPlpiSetIntPar(lpi, SCIP_LPPAR_LPINFO, TRUE) );

If you want to clear the current starting basis, you can use

    SCIP_CALL( SCIPlpiClearState(lpi) );

or specify starting from scratch by

    SCIP_CALL( SCIPlpiSetIntPar(lpi, SCIP_LPPAR_FROMSCRATCH, TRUE) );

The latter is permanent, i.e. all future LPs will be solved from scratch 
until you change this option to FALSE again.

Hope this helps,
Ambros




(This could be deactivated by

Am 07.08.2012 17:21, schrieb Daniel Karch:
> Hi,
>
> I have implemented a Benders decomposition in SCIP via a constraint handler.
> For this I solve the master problem relaxation and give the solution to
> the constraint handler.
> In the constraint handler, I construct a (primal) client LP (as
> SCIP_LPI) and solve its dual by calling
> SCIPlpiSolveDual. If the dual is unbounded, I use the farkas certificate
> to generate a
> violated constraint that is then added to the master.
> Now, every time I call the constraint handler's separation routine, I
> change the primal client LP
> with SCIPlpiChgSides. Therefore, in the dual only the objective function
> changes.
> When I now call SCIPlpiSolveDual, will the LP be solved from scratch, or
> is SCIP/CPlex
> "smart enough" to do a warm start?
> In the documentation I found a mysterious bit about a
> "startFinishOptions" field, is this related?
> How do I use it?
>
> Thanks and best regards,
>
>    Daniel
>
>
> P.S.: I am using CPlex 12.4 as LP solver.
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip
>

-- 
____________________________________________________________
Ambros M. Gleixner
Zuse Institute Berlin - Matheon - Berlin Mathematical School
http://www.zib.de/gleixner