A few questions. Is there more than one version of mpi on the system?
Is LD_LIBRARY_PATH set correctly (does it need to be?")? Try running with "mpirun --prefix /software/mpi/openmpi/1.4.1/i101011" maybe (see http://www.open-mpi.org/faq/)? It is curious because your first post indicates that both trilinos and mpi4py are working with mpi in as far as they are returning the correct processor number. Try using the "nm" command if available on "/software/mpi/openmpi/1.4.1/i101011/lib/openmpi/mca_btl_ofud" and also the "ldd" command can be useful as well. Incidentally, what is "/software/mpi/openmpi/1.4.1/i101011/lib/openmpi/mca_btl_ofu"? If it is executable, maybe you can run it under ltrace or strace to see what is missing to get some clues. Hope this helps in some way. Cheers On Wed, Jan 4, 2012 at 3:47 PM, Ferenc Tasnadi <[email protected]> wrote: > Dear Daniel, > > I tried the diffusion/mesh1D.py with trilinos, what looks working for me. > Or? Here is what I got. > Of course, I commented out the viewer lines. > > $ /software/mpi/openmpi/1.4.1/i101011/bin/mpirun -np 2 python2.7 mesh1D.py > --trilinos > > > [matter1: diffusion]$ /software/mpi/openmpi/1.4.1/i101011/bin/mpirun -np 2 > python2.7 mesh1D.py --trilinos > [matter1:00445] mca: base: component_find: unable to open > /software/mpi/openmpi/1.4.1/i101011/lib/openmpi/mca_btl_ofud: perhaps a > missing symbol, or compiled for a different version of Open MPI? (ignored) > [matter1:00446] mca: base: component_find: unable to open > /software/mpi/openmpi/1.4.1/i101011/lib/openmpi/mca_btl_ofud: perhaps a > missing symbol, or compiled for a different version of Open MPI? (ignored) > [matter1:00445] mca: base: component_find: unable to open > /software/mpi/openmpi/1.4.1/i101011/lib/openmpi/mca_btl_openib: perhaps a > missing symbol, or compiled for a different version of Open MPI? (ignored) > [matter1:00446] mca: base: component_find: unable to open > /software/mpi/openmpi/1.4.1/i101011/lib/openmpi/mca_btl_openib: perhaps a > missing symbol, or compiled for a different version of Open MPI? (ignored) > True > True > Explicit transient diffusion. Press <return> to proceed... > True > True > Implicit transient diffusion. Press <return> to proceed... > True > True > Crank-Nicholson transient diffusion. Press <return> to proceed... > True > True > Implicit steady-state diffusion. Press <return> to proceed... > Time-dependent boundary condition. Press <return> to proceed... > True > True > True > True > [matter1: diffusion]$ > > > Best wishes, > Ferenc > > > > > > On Wed, 4 Jan 2012 14:21:51 -0500 > Daniel Wheeler <[email protected]> wrote: > > > On Wed, Jan 4, 2012 at 12:33 PM, Ferenc Tasnadi <[email protected]> > wrote: > > > > > Then with 8 cores on 1 node exporting FIPY_SOLVERS=Trilinos" I got > after a > > > long run: > > > 2 total processes killed (some possibly by mpirun during cleanup) > > > mpirun noticed that process rank 1 with PID 30180 on node m466 exited > on > > > signal 11 (Segmentation fault). > > > > > > > Can you post all the output from running exactly (not setting > FIPY_SOLVERS, > > use "--trilinos" instead): > > > > $ mpirun -np 16 python examples/diffusion/mesh20x20.py --trilinos > > > > with the viewers commented out and attach the output in a file from > running: > > > > $ mpirun -np 2 python setup test --trilinos > > > > -- > > Daniel Wheeler > _______________________________________________ > fipy mailing list > [email protected] > http://www.ctcms.nist.gov/fipy > [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ] > -- Daniel Wheeler
_______________________________________________ fipy mailing list [email protected] http://www.ctcms.nist.gov/fipy [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]
