[Ipopt] Ipopt Error

Andreas Waechter andreasw at watson.ibm.com
Tue May 3 22:19:29 EDT 2011


Hi Darek,

This is a Fortran runtime error.  There is not string "ICMP" in the Ipopt 
code itself (and Ipopt is written in C++ anyway), but "grep" shows that 
there is an array ICMP in the MUMPS package.

You might have an integer overflow issue, maybe your problem is too large, 
maybe you can compile with 64bit integers, and maybe it helps if you try a 
different linear solver.

However, if your problem is not large, you might be providing bad data to 
Ipopt, e.g., in the position of the nonzeros.  In that case, you may want 
to just print out the values conmputed by the evaluation methods (such as 
eval_jac_g and eval_h) before you return them to Ipopt and look at them 
carefully.  Or you may turn the Ipopt print level very high, and look at 
the matrices as Ipopt sees them.

Hope this helps,

Andreas

On Mon, 2 May 2011 dlle at student.dtu.dk wrote:

> Hi Everybody!
> I am encountering problem when coding my NLP in Ipot and I am getting the
> following message in the visual studio console after solver starts
> calculations:
> "forrtl: severe (408): fort : (3): Subscirpt #1 of the array ICMP has
> value -842150450, which is less than lower bound of 1".
> Seems like i did not initialised something, but i have no idea what array
> ICMP is.
> Regards
> Darek
>
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt
>
>



More information about the Ipopt mailing list