What J-P is suggesting is the following:
Have a
std_cxx11::shared_pointer<TrilinosWrappers::SolverDirect> solver
member in your class, then
if (!solver) {
solver = std_cxx11::shared_pointer<TrilinosWrappers::SolverDirect
direct>(new TrilinosWrappers::SolverDirect direct(…) );
solver->initialize (mass_matrix);
}
this will be done only once, and will be destroyed correctly, avoiding MPI
issues on exit.
L.
> On 24 Jun 2016, at 9:10, Jean-Paul Pelteret <[email protected]> wrote:
>
> Praveen,
>
> I think that the problem is that you have an object that contains MPI data
> (like a MPI communicator) that is still in use/existence at the time that
> MPI_FINALIZE is called (generally by the Utilities::MPI::MPI_InitFinalize)
> object one creates in the main function. You need to ensure that your Solver
> object is destroyed before the MPI_InitFinalize object goes out of scope.
> What you could do is have a (smart-)pointer to your solver in your main
> class, and create and initialise it when (first_time == true), and then it
> gets removed when the class destructor is called.
>
> J-P
>
> On Friday, June 24, 2016 at 8:38:59 AM UTC+2, Praveen C wrote:
> Dear all
>
> I am using the new TrilinosWrappers::SolverDirect which allows two step
> solution: initialize and solve. Since I want to calculate LU decomposition
> only once, I do something like this
>
> static int first_time = 1;
>
> static TrilinosWrappers::SolverDirect::AdditionalData data (false,
> "Amesos_Mumps");
> static SolverControl solver_control (1, 0);
> static TrilinosWrappers::SolverDirect direct (solver_control, data);
>
> // If it is first time, compute LU decomposition
> if(first_time)
> {
> pcout << "Performing LU decomposition\n";
> direct.initialize (mass_matrix);
> first_time = 0;
> }
>
> // solve for ax
> {
> TrilinosWrappers::MPI::Vector tmp (locally_owned_dofs,
> mpi_communicator);
> direct.solve (tmp, rhs_ax);
> constraints.distribute (tmp);
> ax = tmp;
> }
>
> This works but when the program finishes, I get this message.
>
>
> *** The MPI_Comm_f2c() function was called after MPI_FINALIZE was invoked.
> *** This is disallowed by the MPI standard.
> *** Your MPI job will now abort.
> [(null):4853] Local abort after MPI_FINALIZE completed successfully; not able
> to aggregate error messages, and not able to guarantee that all other
> processes were killed!
> *** The MPI_Comm_f2c() function was called after MPI_FINALIZE was invoked.
> *** This is disallowed by the MPI standard.
> *** Your MPI job will now abort.
> [(null):4854] Local abort after MPI_FINALIZE completed successfully; not able
> to aggregate error messages, and not able to guarantee that all other
> processes were killed!
> -------------------------------------------------------
> Primary job terminated normally, but 1 process returned
> a non-zero exit code.. Per user-direction, the job has been aborted.
> -------------------------------------------------------
> --------------------------------------------------------------------------
> mpirun detected that one or more processes exited with non-zero status, thus
> causing
> the job to be terminated. The first process to do so was:
>
> Process name: [[14617,1],0]
> Exit code: 1
> —————————————————————————————————————
>
> Removing “static” for the solver and recomputing it everytime gets rid of
> this error. So I am guessing it is the “static” qualifier that is causing
> problem. How can I fix this issue ?
>
> Thanks
> praveen
>
>
> --
> The deal.II project is located at http://www.dealii.org/
> For mailing list/forum options, see
> https://groups.google.com/d/forum/dealii?hl=en
> ---
> You received this message because you are subscribed to the Google Groups
> "deal.II User Group" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> For more options, visit https://groups.google.com/d/optout.
--
The deal.II project is located at http://www.dealii.org/
For mailing list/forum options, see
https://groups.google.com/d/forum/dealii?hl=en
---
You received this message because you are subscribed to the Google Groups
"deal.II User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
