[Ipopt] Error Message: Restoration phase in the restoration phase failed

Andreas Waechter andreasw at watson.ibm.com
Wed Jan 26 16:30:15 EST 2011


Hi,

It appears that the linear solver (you seem to be using MUMPS) has trouble 
solving the linear system during the step computation (running out of 
memory?).  You could try to use a different linear solver (e.g., Pardiso), 
or you can see if setting the option

mumps_mem_percent

to something less than its default 1000 (e.g., 200 or even less) avoids 
the memory problem in MUMPS.  A 2000 variable problem is not necessarily 
large, but maybe you are trying to solve a problem with dense derivative 
matrices?  (If only the Lagrangian Hessian is dense but the constraints 
Jacobian is sparse, you might be able to speeup up things by using the 
limited-memory Hessian approximation option...)

I hope this helps,

Andreas

On Wed, 26 Jan 2011, Pratim Vakish wrote:

>
>
>
> Hello,
> I am solving large (2000 decision variables) quadratic optimization problems and I obtain error message copied below. For smaller instances (1000 var), Ipopt converges.
>
> Could you please let me know what this error message means and how I could avoid it?
> WARNING: Problem in step computation; switching to emergency
> mode.
>
>   1r9.9999900e-003
> 1.91e+001 9.99e+002   1.3 0.00e+000    -
> 0.00e+000 0.00e+000R  1
>
> MUMPS returned INFO(1) =-13 - out or memory.Restoration phase is called at point that is almost
> feasible,  with constraint
> violation 1.734723e-018. Abort.
>
> Restoration phase in the restoration phase failed.
>
>
> Thanks,
>
> Pratim
>
>



More information about the Ipopt mailing list