0

Some time ago we wrote our own mpi wrapper for python. Everything worked fine on mac after I recently upgraded to:

openmpi-default @1.7.3_2+gcc48 (active)
openmpi-devel-default @1.9a1_30433+gcc48 (active)
openmpi-gcc45 @1.7.3_2+fortran (active)

Everything with the code below was running fine, and still does on linux, albeit versions may be different. However, here's the code. The functions are essentially wrappers.

#!/usr/bin/env python


import mpilib

if not mpilib.isInitialised():
    print 'af'
mpilib.init()

Calling init() gives the following error:

mpiexec -c 3 test.py

af
af
af
[localhost:02479] [[29887,0],0] mca_oob_tcp_recv_handler: invalid message type: 1959854080
[localhost.org:02479] [[29887,0],0] mca_oob_tcp_recv_handler: invalid message type: 1958674432
-------------------------------------------------------
Primary job  terminated normally, but 1 process returned
a non-zero exit code.. Per user-direction, the job has been aborted.
-------------------------------------------------------
--------------------------------------------------------------------------
mpiexec detected that one or more processes exited with non-zero status, thus causing
the job to be terminated. The first process to do so was:

Process name: [[29887,1],0]
Exit code:    1

The message type is not reproducible and appears to be a not initialized memory segment. Ideas anyone?

Thanks, EL

El Dude
  • 5,328
  • 11
  • 54
  • 101

0 Answers0