Problem resolved when I built an mpi enabled python through mpi4py:
https://groups.google.com/forum/#!topic/mpi4py/Puv4BHqQ4qY


On Thu, Apr 17, 2014 at 1:52 PM, Mike Welland <[email protected]> wrote:

> Hi all,
>
> I built fenics on our cluster and got to the point where I can sometimes
> completely run the demo_hyperelasticity.py, but only once!
>
> Here is the rundown:
>
> 1.  In the python2.7/site-packages/dolfin directory I run:
> rm `find ./ -name '*.pyc'` -rf
> 2. In another terminal I run
> python demo_hyperelasticity.py
> and it works all the way through
> 3. In the same terminal I repeat the command exactly, and it immediately
> crashes:
> Fatal error in PMPI_Init_thread: Internal MPI error!
>
> From that, it looks like it is in the python compilation of the pyc files
> that the problem lies. Has anyone seen this before?
>
> Possibly related:
> I still get the
> python:96004 terminated with signal 7 at PC=37f720c380 SP=7fff17027758.
>  Backtrace:
> /lib64/libpthread.so.0(pthread_spin_lock+0x0)[0x37f720c380]
> ...
>
> error but it goes away if I just repeat the commands. It seems to come
> more often if I delete ~/.instant but I can't reliably reproduce it.
>
> Thanks,
> Mike
>
>
>
>
_______________________________________________
fenics-support mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics-support

Reply via email to