This error message is indicating that MPI_Comm_f2c is being invoked before MPI_INIT.

Can you see if that is happening?

If it helps, you can run with the MCA parameter mpi_abort_print_stack set to 1. This will print the stack when we abort. Or run with the MCA parameter mpi_abort_delay set to -1; this will print the abort message but then spin forever (i.e., not abort), which allows you to attach a debugger to the MPI processes and poke around to figure out what went wrong.

See http://www.open-mpi.org/faq/?category=debugging#debug-ompi- controls for a full list of the MCA parameters that we have to help debugging.


On Jun 22, 2009, at 4:14 PM, Si Hammond wrote:

Hi everyone,

Another little mystery for you all!

I have a relatively small MPI Fortran code which I can compile successfully with Intel and PGI compiled OpenMPI 1.3.0 (and 1.2.5). No problems with this.

On a separate machine I have a GNU Gfortran 4.3 OpenMPI 1.3.0 installation and I get these errors immediately after the code starts running:

*** An error occurred in MPI_Comm_f2c
*** before MPI was initialized
*** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)
[***:19386] Abort before MPI_INIT completed successfully; not able to guarantee that all other processes were killed!
*** An error occurred in MPI_Comm_f2c
*** before MPI was initialized
*** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)
[***:13269] Abort before MPI_INIT completed successfully; not able to guarantee that all other processes were killed!
*** An error occurred in MPI_Comm_f2c
*** before MPI was initialized
*** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)
<etc>


So, it looks like there might be a bug with GFortran or do you think this is OpenMPI?

Any help you have would be appreciated, I can't seem to find much on this using Google.




Si Hammond

Performance Modelling, Analysis and Optimisation Team
High Performance Systems Group
Department of Computer Science
University of Warwick, CV4 7AL, UK





_______________________________________________
users mailing list
us...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/users


--
Jeff Squyres
Cisco Systems

Reply via email to