[MOAB-dev] [Fathom] Problem with mbparallelcomm_test?

[Dmitry Karpeev]

yes, but I'm just at a loss at the difference between mbconvet and
mbparallelcomm_test.
the code is very straightforward and dispatches to load_file in either case.
I thought there may be a problem with the file, but mbconvert appears
to read it fine.
I stripped down mbconvert to a single read call with timing around it
and a few basic arguments.
Hopefully that will work as a straightforward benchmarking tool, but
it still doesn't explain this odd
discrepancy.

Dmitry.

On Fri, Apr 2, 2010 at 7:37 PM,  <tautges at mcs.anl.gov> wrote:
> Bcast puts the whole mesh on every proc, so no shared ents. I think this is
> documented in the v4 ug now.
>
> - tim
>
>
> On Apr 2, 2010, at 4:01 PM, Dmitry Karpeev <karpeev at mcs.anl.gov> wrote:
>
>> I ran mbconvert and mbparallelcomm_test on blocks-64.h5m and
>> tjunc6RIB.h5m using 4 procs on cosmea.
>> It appears that mbconvert has no problem reading these files with
>> PARALLEL=BCAST, but mbparallelcomm_test isn't doing so well.
>> I'm not sure what the difference between the two is (I suppose I
>> should go and sift through the source code), but perhaps mbconvert
>> doesn't have to
>> resolve shared entities, while mbparallelcomm_test does?
>>
>> The mesh files are in ~karpeev/fathom/moab/data/blocks-64.h5m and
>> ~karpeev/fathom/moab/data/tjunc6/tjunc6RIB_16.h5m on MCS machines or
>> on cosmea.
>> Any ideas about what's going on?
>> Thanks.
>> Dmitry.
>>
>> Here's the output:
>>
>> ============================================================================================================================
>> blocks-64/mbconvert
>>
>> -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>> --- Running LD_PRELOAD= mpiexec -machinefile mpi.nodes.17204 -n 4
>> /home/karpeev/fathom/moab/data/../unstable/build/tools/mbconvert  '-O'
>> 'PARALLEL=BCAST' '-O' 'PARTITION=MATERIAL_SET' '-f' 'VTK'
>> 'blocks-64.h5m' '/dev/null' 2>&1 1>
>> /home/karpeev/fathom/moab/data/./mbconvert_runs/bcast/blocks-64.4.out
>> Read "blocks-64.h5m"
>> Wrote "/dev/null"
>> Read "blocks-64.h5m"
>> Wrote "/dev/null"
>> Read "Read "blocks-64.h5m"
>> blocks-64.h5m"
>> Wrote "/dev/null"
>> Wrote "/dev/null"
>> --- Done
>>
>> -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>> blocks-64/mbparallelcomm_test
>>
>> -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>>
>> --- Running LD_PRELOAD= mpiexec -machinefile mpi.nodes.17589 -n 4
>>
>> /home/karpeev/fathom/moab/data/../unstable/build/test/parallel/mbparallelcomm_test
>> '-3' '3' '1' 'tjunc6/blocks-64.h5m' 'MATERIAL_SET' 2>&1 1>
>>
>> /home/karpeev/fathom/moab/data/./mbparallelcomm_test_runs/bcast/blocks-64.4.out
>> Couldn't read mesh; error message:
>> Failed in step PARALLEL READ
>>
>>
>> application called MPI_Abort(MPI_COMM_WORLD, 0) - process 0--- Done
>>
>> -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>> tjunc6RIB_16/mbconvert
>>
>> -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>> --- Running LD_PRELOAD= mpiexec -machinefile mpi.nodes.17340 -n 4
>> /home/karpeev/fathom/moab/data/../unstable/build/tools/mbconvert  '-O'
>> 'PARALLEL=BCAST' '-O' 'PARTITION=PARALLEL_PARTITION' '-f' 'VTK'
>> 'tjunc6/tjunc6RIB_16.h5m' '/dev/null' 2>&1 1>
>> /home/karpeev/fathom/moab/data/./mbconvert_runs/bcast/tjunc6RIB_16.4.out
>> Read "tjunc6/tjunc6RIB_16.h5m"
>> Read "tjunc6/tjunc6RIB_16.h5m"
>> Read "tjunc6/tjunc6RIB_16.h5m"
>> Read "tjunc6/tjunc6RIB_16.h5m"
>> Wrote "/dev/null"
>> Wrote "/dev/null"
>> Wrote "/dev/null"
>> Wrote "/dev/null"
>> --- Done
>>
>> -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>> tjunc6RIB_16/mbparallelcomm_test
>>
>> -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>> --- Running LD_PRELOAD= mpiexec -machinefile mpi.nodes.17453 -n 4
>>
>> /home/karpeev/fathom/moab/data/../unstable/build/test/parallel/mbparallelcomm_test
>> '-3' '3' '1' 'tjunc6/tjunc6RIB_16.h5m' 'PARALLEL_PARTITION' 2>&1 1>
>>
>> /home/karpeev/fathom/moab/data/./mbparallelcomm_test_runs/bcast/tjunc6RIB_16.4.out
>> Couldn't read mesh; error message:
>> Failed in step PARALLEL RESOLVE_SHARED_ENTS
>>
>>
>> application called MPI_Abort(MPI_COMM_WORLD, 0) - process 0--- Done
>> _______________________________________________
>> Fathom mailing list
>> Fathom at lists.mcs.anl.gov
>> https://lists.mcs.anl.gov/mailman/listinfo/fathom
>