I had this kind of problem before, so I already checked that. The wrapper
compiler
and mpirun match.
A very simple testprogram that I used gave me this result:
Total sum: 0. id 1
Total sum: 1014.1131933453868 id 0
Total sum: 0.000
Are you saying that if you build and run test program 1, you see MPI_COMM_WORLD
ranks of 0-(N-1), and if you build and run test program 2, you see N
MPI_COMM_WORLD ranks 0?
I would run ldd on your resulting executables and verify that they are linking
against the libmpi.so that they should.
O
Dear all,
In Java bindings I can't use MPI_STATUS_IGNORE in MPI_Waitany nor
MPI_Testany. I must use a MPI_Status in order to avoid a SIGSEGV error.
I searched for this issue but I don't found any error report.
Should I report a bug?
Regards,
Oscar
--
Yes, please -- let's track this so that it doesn't get forgotten. Mark it as
critical for 1.8. It's too late for 1.7.5, sorry. :-(
On Feb 26, 2014, at 2:25 PM, Oscar Vega-Gisbert wrote:
> Dear all,
>
> In Java bindings I can't use MPI_STATUS_IGNORE in MPI_Waitany nor
> MPI_Testany. I must
There is an OMPI_RML_TAG_SHMEM in ompi/mca/rte/rte.h on the v1.7 branch that
does not exist in the trunk. I cannot find it used anywhere in the v1.7 branch
(or the trunk, obviously).
I'm taking it out as part of a merge conflict on the v1.7 branch, but it
worries me a bit that it is there: it
On Feb 26, 2014, at 2:58 PM, Jeff Squyres (jsquyres) wrote:
> There is an OMPI_RML_TAG_SHMEM in ompi/mca/rte/rte.h on the v1.7 branch that
> does not exist in the trunk. I cannot find it used anywhere in the v1.7
> branch (or the trunk, obviously).
>
> I'm taking it out as part of a merge co
Getting a bunch of these on the trunk:
../../../oshmem/mca/memheap/memheap.h:142:5: warning:
"MCA_oshmem_memheap_DIRECT_CALL" is not defined
Can someone investigate?
Thanks.
--
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_busines
Also getting a bunch of these from clang:
scoll_mpi_ops.c:21:46: warning: must specify at least one argument for '...'
parameter of variadic macro [-Wgnu]
MPI_COLL_VERBOSE(20,"RUNNING MPI BARRIER");
^
./scoll_mpi_debug.h:20:9: note: macro 'MPI
