[SCIP] FiberSCIP continue running after finding a solution that fits best solution

[Stefan Vigerske]

Hi,

I think we found the problem. This patch should help to make it 
terminate when the gap is 0:


--- a/src/ug_scip/scipParaSolver.cpp
+++ b/src/ug_scip/scipParaSolver.cpp
@@ -304,6 +304,7 @@ ScipParaSolver::setRacingParams(
              default:
                 THROW_LOGICAL_ERROR1("invalid nHeuristics");
           }
+         SCIP_CALL_ABORT( SCIPsetIntParam(scip, 
"heuristics/ObjLim/freq", 1) );

           switch( nPresolving )
           {



That is, add
   SCIP_CALL_ABORT( SCIPsetIntParam(scip, "heuristics/ObjLim/freq", 1) );
after the setting of primal heuristic parameters in 
ScipParaSolver::setRacingParams().

Best,
Stefan

On 10/04/2024 12:54, Santiago Garcia wrote:
> Hi,
> 
> I'm using UG framework and FiberSCIP executed from command line to solve a maximization problem. I'm trying to finetune the settings to optimize the time to get a result, and I'm facing a strange situation where the process is still running after finding a solution that fits exactly with the objective value (best solution).
> 
> Please find attached the files with the problem and parameters introduced, and the output.
> 
> If I establish a timelimit that happens before finding a solution that fits the objective, the command returns correctly with the best fit found till the moment, but if I establish a bigger timelimit, and the solution found fits the best possible one, then the process stops with the time limit but it returns an error code.
> 
> I'd really appreciate your support on this issue.
> 
> 
> Santiago García
> 
> 
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip