[ET Trac] [Einstein Toolkit] #1473: Carpet segfaults in Shutdown

[Einstein Toolkit]

#1473: Carpet segfaults in Shutdown
---------------------+------------------------------------------------------
  Reporter:  knarf   |       Owner:  eschnett           
      Type:  defect  |      Status:  new                
  Priority:  minor   |   Milestone:                     
 Component:  Carpet  |     Version:  development version
Resolution:          |    Keywords:                     
---------------------+------------------------------------------------------

Comment (by rhaas):

 I just tried this on bethe (9 cores, ht, intel13). Using 9 OpenMP threads
 and 1 process I get an assertion failure from within hwloc:
 {{{  MPI process 0 on host 0 (process 0 of 1 on this host)
     OpenMP thread 0: PU set L#{0-15} P#{0-15}
     OpenMP thread 1: PU set L#{0-15} P#{0-15}
     OpenMP thread 2: PU set L#{0-15} P#{0-15}
     OpenMP thread 3: PU set L#{0-15} P#{0-15}
     OpenMP thread 4: PU set L#{0-15} P#{0-15}
     OpenMP thread 5: PU set L#{0-15} P#{0-15}
     OpenMP thread 6: PU set L#{0-15} P#{0-15}
     OpenMP thread 7: PU set L#{0-15} P#{0-15}
     OpenMP thread 8: PU set L#{0-15} P#{0-15}
 INFO (hwloc): Setting thread CPU bindings:
 cactus_sim:
 /home/rhaas/ET_trunk/arrangements/ExternalLibraries/hwloc/src/system_topology.cc:525:
 void <unnamed>::set_bindings(hwloc_topology *, const
 <unnamed>::mpi_host_mapping_t &): Assertion `core_num < num_cores' failed.
 }}}
 To run, do:
 {{{
 export
 CCTK_TESTSUITE_RUN_COMMAND="/usr/local/apps/openmpi-1.6-intel-2013.2.146/bin/mpirun
 -np \$nprocs $NICE \$exe \$parfile"
 OMP_NUM_THREADS=9 make sim-testsuite
 }}}
 then select 1 process, and the CarpetWaveToyNewRecover_test_1proc test of
 CarpetIOHDF5. The test passes with 8 threads and 4 threads (and also 16).
 It fails also with 9, 11 threads.

-- 
Ticket URL: <https://trac.einsteintoolkit.org/ticket/1473#comment:2>
Einstein Toolkit <http://einsteintoolkit.org>
The Einstein Toolkit