Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

Subject: Re: [OMPI users] MPI_COMM_WORLD Error
From: Gus Correa (gus_at_[hidden])
Date: 2009-05-26 20:50:38


Hi Trent, list

mtcreekmore_at_[hidden] wrote:
> Thanks for your reply Eugene.
>
>
>
> Here is the result of the mpif90 -showme
>
>
> /state/partition1/apps/intel/fce/10.1.015/bin/ifort
> -I/share/apps/openmpi-1.2.6-intel/include -pthread
> -I/share/apps/openmpi-1.2.6-intel/lib -L/usr/pbs/lib
> -L/share/apps/openmpi-1.2.6-intel/lib -lmpi_f90 -lmpi_f77 -lmpi
> -lopen-rte -lopen-pal -lrt -lpbs -lnuma -ldl -Wl,--export-dynamic -lnsl
> -lutil
>
>
> The mpif.h file I got off the site. I used the Google "site: mcs.anl.gov
> <http://www.mcs.anl.gov/research/projects/mpi/usingmpi/examples/simplempi/main.htm> mpif.h"
> search and placed it into the same directory as the main program.
>

Then you are using the MPICH mpif.h file, not the OpenMPI one,
and the mix is probably the reason for the error.
You can/should remove it from the directory where your pi.f program is.

Actually, if you use the OpenMPI wrappers to compile,
they know exactly where to find the right mpif.h,
i.e., to compile pi.f just do "mpif90 pi.f".

Make sure your PATH is points to the OpenMPI executables (mpif90,
mpiexec, etc) ahead of other flavors of MPI that may be installed on
your system. Or simply use the full path names to the OpenMPI mpif90
and to mpiexec. Inadvertent mix of these executables
from different MPIs is a common source of frustration too.

I hope this helps,
Gus Correa
---------------------------------------------------------------------
Gustavo Correa
Lamont-Doherty Earth Observatory - Columbia University
Palisades, NY, 10964-8000 - USA
---------------------------------------------------------------------

> So if it is the correct one, I am not sure.
>
> Trent
>
>
>
> ------------------------------------------------------------------------
> Date: Tue, 26 May 2009 16:54:02 -0700
> From: Eugene.Loh_at_[hidden]
> To: users_at_[hidden]
> Subject: Re: [OMPI users] MPI_COMM_WORLD Error
>
> mtcreekmore_at_[hidden] <mailto:mtcreekmore_at_[hidden]> wrote:
>
>
> I and new to OpenMPI and have looked for this, but not getting
> anywhere.
>
> I got an example program off this site:
>
> The compute PI in Fortran example.
> http://www.mcs.anl.gov/research/projects/mpi/usingmpi/examples/simplempi/main.htm
>
> Compiled it with "mpif90" and no errors.
> Then used the qsub -I and tried both executing directly and "mpirun"
> and still receive these errors:
>
> [MGT-2-1:02994] *** An error occurred in MPI_Comm_rank
> [MGT-2-1:02994] *** on communicator MPI_COMM_WORLD
> [MGT-2-1:02994] *** MPI_ERR_COMM: invalid communicator
> [MGT-2-1:02994] *** MPI_ERRORS_ARE_FATAL (goodbye)
>
> Any guidance in the right direction would be appreciated.
>
> First, the example works fine for me.
>
> You seem to be bombing in MPI_Comm_rank and before it asks you for any
> input. I guess it's having problems with the MPI_COMM_WORLD
> communicator being passed to MPI_Comm_rank? Maybe it's having problems
> finding the mpif.h include file? It's finding *some* mpif.h include
> file, but presumably not the right one. Make sure the mpif90 and mpirun
> commands are the ones you think you should be getting and, if they are,
> try adding -showme to the mpif90 line to see where it's looking for the
> mpif.h include file.
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> users mailing list
> users_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/users