[Ipopt] Linking Ipopt to MPI code

[Uwe Sander]

  Hi Damien,

thanks for your advice. I am indeed using MUMPS at the moment. I'll to 
throw it out and use another linear solver instead.

Cheers,
Uwe



On 10/12/2010 05:02 PM, Damien wrote:
>    Uwe,
>
> Are you using MUMPS?  That has fake MPI headers for the sequential
> version in IPOPT and that doesn't play nicely with real MPI libraries.
>
> Damien
>
> On 12/10/2010 8:51 AM, Uwe Sander wrote:
>>     Hi folks,
>>
>> I have an existing C++/MPI code and would like to combine it with Ipopt.
>> I have added the necessary inlcude files and libraries to my Makefile,
>> without adding any Ipopt code to my source files yet. Unfortunately, I
>> noticed the following: As soon as I link the ipopt library to my
>> project, the code does not run properly. It still compiles without
>> errors, but on execution I get the error message
>>
>> "MPI_COMM_WORLD rank 0 has terminated without calling MPI_Finalize()".
>>
>> If I compile the same code without the "-lipopt" option for the linker,
>> the program runs fine.
>>
>> I doubt that error arises from the MPI code since I tested this with a
>> very basic program that consisted of only four MPI steps: MPI_Init,
>> MPI_Comm_size, MPI_Comm_rank, and MPI_Finalize. Here, I have the same
>> problem as above.
>>
>> Any ideas on this?
>>
>>
>> Thanks,
>> Uwe
>>
>> _______________________________________________
>> Ipopt mailing list
>> Ipopt at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/ipopt
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt
>