[IPython-user] Initializing MPI
Thu Nov 12 12:32:05 CST 2009
Sorry, but I am horribly busy today. But here is a short answer.
* Yes, you can use anything you want to initialize MPI.
* If you want the engine ids to match the MPI ranks, you will need to use
an ipython config file to add support for your method of initializing MPI.
See IPython/kernel/scripts/ipengine.py and IPython/kernel/config/
for details and examples.
Note: I have completely refactored how ipcluster works in a branch. It will
much easier to config all of this in the near future. But the overall idea
be the same.
Things will be better for me next week, but please don't hesitate to ask
On Thu, Nov 12, 2009 at 8:47 AM, Stefan Seefeld <firstname.lastname@example.org>wrote:
> I have been developing a C++ library which is using MPI internally. I
> have also written Python wrappers for it, so I can call into it from a
> normal Python script.
> Now I'd like to use the library via ipython. Specifically, I'e like to
> be able to use parallel ipython (using ipcluster), load my extension
> module into all engines, then run it. I have been trying to figure out
> how to initialize MPI to make that work. I have read
> which suggests to use mpi4py.
> I have done that, and all appears to be working fine. However, given
> that I don't actually need mpi4py for anything but initialization, I
> wonder whether I can't simply call MPI_Init() myself (as I would do when
> using my library in a stand-alone application).
> Thus my question: Is there a way to initialize MPI that does not involve
> mpi4py ? (The above document states:
> "We highly recommend just using mpi4py as it takes care of most of these
> problems. If you want to do something different, let us know and we can
> help you get started.",
> which I'd like to take advantage now. :-) )
> Many thanks,
> ...ich hab' noch einen Koffer in Berlin...
> IPython-user mailing list
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the IPython-user