[Scipy-tickets] [SciPy] #1835: mpiexec with numpy causes segfault

SciPy Trac scipy-tickets@scipy....
Fri Feb 1 15:17:44 CST 2013


#1835: mpiexec with numpy causes segfault
-----------------------+----------------------------------------------------
 Reporter:  freechus9  |       Owner:  somebody   
     Type:  defect     |      Status:  new        
 Priority:  normal     |   Milestone:  Unscheduled
Component:  Other      |     Version:  0.11.0     
 Keywords:             |  
-----------------------+----------------------------------------------------
 Hello, I have been scratching my head on this problem all day.

 I am installing numpy,scipy, and mpi4py from source on a cluster, and I
 have done this before on a different account.  Everything works peachy
 over there, so I did the exact same thing for the new user.

 Here is what happens:

 If I import numpy or scipy on a single core, everything works fine.
 However, as soon as I do mpiexec -n 4 python test.py I get an immediate
 segfault.  In addition, if I try to re-import numpy with a single core, I
 still get a segfault.  Even more interestingly, if I wait a small amount
 of time, the problem goes away for single core.  However, repeating the
 same process yeilds the same results.

 Not sure if this is relevant, but I built numpy and scipy against the same
 Openblas shared .so file that I compiled with the other user.  Meaning I
 just copied over all .so and .a files..it is the same machine so it must
 be the same architecture..

 Thanks.

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1835>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list