On Mar 28, 2009, at 5:31 AM, François PELLEGRINI wrote:

  MPI_Comm_size (MPI_COMM_WORLD, &procglbnbr);

  MPI_Comm_rank (MPI_COMM_WORLD, &proclocnum);

  dataloctab = malloc (2 * (procglbnbr + 1) * sizeof (int));
  dataglbtab = dataloctab + 2;


Why +2?  Shouldn't it be +procglbnbr?

3) Messages
===========

In addition to the "memcpy" message, I also get a bunch
of strange messages. Some excerpts :

==30478== Syscall param sched_setaffinity(mask) points to unaddressable byte(s)
==30478==    at 0x434F1D7: syscall (in /lib/i686/libc-2.8.so)
==30478==    by 0x420B1A4: opal_paffinity_linux_plpa_api_probe_init
(plpa_api_probe.c:43)
==30478== by 0x420B6E3: opal_paffinity_linux_plpa_init (plpa_runtime.c:36) ==30478== by 0x420C43C: opal_paffinity_linux_plpa_have_topology_information
(plpa_map.c:501)
==30478== by 0x420ABB2: linux_module_init (paffinity_linux_module.c:119) ==30478== by 0x41D6039: opal_paffinity_base_select (paffinity_base_select.c:64)
==30478==    by 0x41A187E: opal_init (opal_init.c:292)
==30478==    by 0x412D010: orte_init (orte_init.c:76)
==30478==    by 0x406784F: ompi_mpi_init (ompi_mpi_init.c:342)
==30478==    by 0x40A4282: PMPI_Init (pinit.c:80)
==30478==    by 0x8048733: main (in /tmp/brol)
==30478==  Address 0x0 is not stack'd, malloc'd or (recently) free'd


Don't worry about this one. It's an internal test in Open MPI's processor affinity where it is probing the kernel to see what kind of affinity support is available (the test is designed to pass a NULL down -- it's safe). If you configure Open MPI 1.3.2 and later --with- valgrind, this warning will go away.

==30477== Conditional jump or move depends on uninitialised value(s)
==30477==    at 0x480AA03: mca_mpool_sm_alloc (mpool_sm_module.c:79)
==30477==    by 0x48491B3: mpool_calloc (btl_sm.c:108)
==30477==    by 0x4849BDC: sm_btl_first_time_init (btl_sm.c:307)
==30477==    by 0x484A1E2: mca_btl_sm_add_procs (btl_sm.c:484)
==30477==    by 0x483F433: mca_bml_r2_add_procs (bml_r2.c:206)
==30477==    by 0x481E7AE: mca_pml_ob1_add_procs (pml_ob1.c:308)
==30477==    by 0x4067F4E: ompi_mpi_init (ompi_mpi_init.c:667)
==30477==    by 0x40A4282: PMPI_Init (pinit.c:80)
==30477==    by 0x8048733: main (in /tmp/brol)



I *believe* that this bug has been fixed since 1.3.1.

--
Jeff Squyres
Cisco Systems


Reply via email to