[IPython-dev] visualization with parallel ipython

Prabhu Ramachandran prabhu at aero.iitb.ac.in
Tue May 8 13:33:01 EDT 2007


>>>>> "Doug" == Doug Jones <dfj225 at gmail.com> writes:

    Doug> Are there any users of IPython that currently do
    Doug> visualization of parallel data sets? If so, do you know what
    Doug> tools they find useful?

I don't do any parallel visualization but do know that VTK does
support parallel processing via MPI.  It is probably not very Pythonic
an approach though.  Kitware also has a powerful open source tool
called ParaView that does parallel visualization.  While paraview is
very powerful I am not sure that it is very Python friendly in the
sense that it integrates with numpy/ipython very well.

    Doug> It seems that in serial environments pylab and mayavi are
    Doug> popular choices, so it might be likely that users want to
    Doug> use these packages in a parallel environment.

    Doug> Forgive my ignorance, but I am not really familiar with
    Doug> pylab or mayavi, but I'm hoping a flow similar to this would
    Doug> be feasible for these packages.

I think this might be feasible but it is a good idea to look at what
VTK/ParaView do and learn from that before doing any work in this
direction.  If you have mayavi related questions in this direction
answered please post to enthought-dev at mail.enthought.com or
mayavi-users at lists.sourceforge.net and I'll try and respond.

    Doug> One IPython1 node collects the data necessary for the plot.
    Doug> This node uses a visualization package to produce the plot
    Doug> or visualization (a graph or geometry).  This resulting
    Doug> visualization is serialized and sent to the client.  The
    Doug> client can then view and further manipulate using a plotting
    Doug> tool on their desktop machine.

Yes, this is what I think is done in ParaView.  2D is a lot easier but
3D is harder and AFAIK ParaView uses a library called IceT to do the
composting of the final image that is rendered on screen.

HTH.

cheers,
prabhu



More information about the IPython-dev mailing list