Anton,

> There are ten inputs originally collected for GLPK 4.55.
> Three of them do not confuse GLPK since 4.56, at least for one
> hour horizon. Another two invariably force the solver to fail.
> 
> >From 4.55, 4.56 to 4.59, 4.59.2, the set of errors has changed.
> Now it includes, besides locking in dual simplex,
>    Assertion failed: tcol[csa->p] != 0.0
>    Error detected in file ..\src\simplex\spydual.c at line ...
> and
>    glp_intopt: cannot solve current LP relaxation.
> 

Thank you very much for your evaluation of glpk. Your models have some
unusual numerical properties that helps me to reveal some rare cases in
the simplex algorithms.

I added another feature to the glpk dual simplex solver, which is the
perturbation of objective coefficients used to avoid numerical
instability and cycling.

I will inform you about a new release of glpk.


Best regards,

Andrew Makhorin



_______________________________________________
Bug-glpk mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/bug-glpk

Reply via email to