reassign 625832 libopenmpi1.3
thanks
Since hangs in non-python programs it has to do with a generic
install/configuration of openmpi, thus reassigning to openmpi.
On Sun, 08 May 2011, Juha Jäykkä wrote:
> Let's see...
> > python -c 'import mpi4py.MPI'
> No use, hangs similarly.
> > 2. where does it hang? (according to strace or may be gdb)?
> Strace and gdb trace attached. Here are the relevant parts of ps and netstat
> outpot, too.
> orted --hnp --set-sid --report-uri 7 --singleton-died-pipe 8
> tcp 0 0 *:60561 *:* LISTEN
>
> 6238/orted
> tcp6 0 0 [::]:34267 [::]:* LISTEN
>
> 6238/orted
> > 3. does any other (non-Python) application built against openmpi ran
> > with opterun hangs? (i.e. is it configured properly, etc)
> Hangs in exaxtly the same place:
> #0 0x00007ffff690e668 in __poll (fds=0x629ac0, nfds=4, timeout=<value
> optimized out>)
> at ../sysdeps/unix/sysv/linux/poll.c:83
> #1 0x00007ffff76a3921 in ?? () from /usr/lib/libopen-pal.so.0
> #2 0x00007ffff76a28cf in ?? () from /usr/lib/libopen-pal.so.0
> #3 0x00007ffff76970b1 in opal_progress () from /usr/lib/libopen-pal.so.0
> #4 0x00007ffff5a2c7e5 in ?? () from
> /usr/lib/openmpi/lib/openmpi/mca_rml_oob.so
> #5 0x00007ffff5a2ccb0 in ?? () from
> /usr/lib/openmpi/lib/openmpi/mca_rml_oob.so
> #6 0x00007ffff7915986 in orte_routed_base_register_sync () from
> /usr/lib/libopen-rte.so.0
> #7 0x00007ffff60384ce in ?? () from
> /usr/lib/openmpi/lib/openmpi/mca_routed_binomial.so
> #8 0x00007ffff7902652 in orte_ess_base_app_setup () from /usr/lib/libopen-
> rte.so.0
> #9 0x00007ffff5c30fa1 in ?? () from
> /usr/lib/openmpi/lib/openmpi/mca_ess_singleton.so
> #10 0x00007ffff78e9ba3 in orte_init () from /usr/lib/libopen-rte.so.0
> #11 0x00007ffff7b62397 in ?? () from /usr/lib/libmpi.so.0
> #12 0x00007ffff7b833e0 in PMPI_Init () from /usr/lib/libmpi.so.0
> #13 0x0000000000400916 in main (argc=1, argv=0x7fffffffddc8) at hello.c:5
> That result is independent of number of ranks and whether I start the program
> with or without orterun. The hello.c is very short:
> #include <mpi.h>
> #include <stdio.h>
> int main(int argc, char **argv) {
> int rc, id;
> rc=MPI_Init(&argc, &argv);
> rc=MPI_Comm_rank(MPI_COMM_WORLD, &id);
> printf("My id = %i\n", id);
> MPI_Finalize();
> }
> And, as you see from the backtrace, it hangs in MPI_Init().
> Does this mean that the problem is not actually mpi4py, but openmpi (or some
> such) instead?
> > altogether it might be simply a generic misconfiguration of mpi. If we
> I do not believe this. I never configured openmpi AT ALL, every single
> openmpi
> setting is pristine, as they came out of the Debian package. I never had to
> touch anything - which is precisely the reason I started using openmpi all
> those years ago, when mpich needed tweaking to get it running at all.
> Cheers,
> Juha
--
=------------------------------------------------------------------=
Keep in touch www.onerussian.com
Yaroslav Halchenko www.ohloh.net/accounts/yarikoptic
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
