[Ipopt] Ipopt terminates

[Andreas Waechter]

Hi

Well, it might also just be a problem with Lapack and matlab and 64bit and 
funny integer types.  The error message is from LAPACK, and given that the 
signature is

            SUBROUTINE DPOTRF( UPLO, N, A, LDA, INFO )

            CHARACTER      UPLO

            INTEGER        INFO, LDA, N

            DOUBLE         PRECISION A( LDA, * )

and therefore LDA seems to have an illegal value, it might be that Fortran 
and C integer types do not match.

This might be related to a recent discussion on the Ipopt mailing list:

http://list.coin-or.org/pipermail/ipopt/2011-March/002376.html

And I have also seen a similar problem in a different context (not with 
matlab but also related to different integer types).  For this 
specific case, I added some code in Ipopt/src/LinAlg/IpLapack.cpp (look 
for the macro FUNNY_LAPACK_FINT).  I suggest you try to activate this code 
by configuring with

ADD_CXXFLAGS='-DFUNNY_LAPACK_FINT'

and recompiling everything (after 'make clean').  Not sure if that will 
help, but it might be worth a try...

Andreas

On Mon, 14 Mar 2011, Stefan Vigerske wrote:

> Hi,
>
> it probably means that the jacobian or hessian contains values like nan
> or inf, which may be a result from a nondifferentiability of one of your
> functions in the current point.
>
> There is the Ipopt option check_derivatives_for_naninf
> http://www.coin-or.org/Ipopt/documentation/node89.html
> which may help to verify this.
>
> Stefan
>
>> I'm using ipopt with the matlab interface. So far everything went fine
>> and I had no problems. But now I have a problem, that can be solve with
>> matlab's fmincon but not with ipopt. When I run the problem with ipopt I
>> get the following error message
>>
>>  ** On entry to DPOTRF parameter number  4 had an illegal value
>>
>> I do not understand what this means.
>>
>> Paul
>>
>> _______________________________________________
>> Ipopt mailing list
>> Ipopt at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/ipopt
>>
>
>
> -- 
> Stefan Vigerske
> Humboldt University Berlin, Numerical Mathematics
> http://www.math.hu-berlin.de/~stefan
>
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt
>
>