[Scip] Different results using SOPlex and CPLEX

[Kati Wolter]

Hi Martin,


Am 19.01.2010 10:38, schrieb Martin Bergner:
> Hi all,
>
> when I compile my program with OPT=dbg LPS=cpx, it works flawlessly.
> However, compiling with soplex, apparently setting all of my variables
> to the value 0 is a valid solution. Since I have set partitioning
> constraints in my problem (\sum x_i = 1), this can not be right. 
>
>   
I ran your lp-file with SCIP 1.2.0 in debug mode using first CPLEX 12.10
and afterwards SoPlex 1.4.2 as underlying LP solver. In both cases, the
same optimal solution was returned. It has an optimal solution value of
755.381833333333 and for each order_* constraint, exactly one of the
included variables has value 1, i.e., it is not the infeasible all zero
solution.

> This problem exists with scip 1.2.0. Exporting the problem as lp and
> solving it with commandline cplex works. With scip in the commandline
> it works with cplex and not with soplex.
>
>   

Therefore, could you please send us a log-file for both of your runs
(scip+cplex and scip+soplex) including problem and optimization
statistics (in the shell, type "display statistics" after your problem
has been solved).


Best regards,
Kati