Interest in testing ParaView with a custom MPI plugin - support needed

Development
1

Hi all,

I am working on an MPI based extension library, and I would really love to test it within ParaView, since I think it would be a very interesting real-world application.
However, to do so, I’d need (and love) your guidance, since I never used the tool, nor I know how the code is structured.

For that, I have a couple of questions to start with:

  1. First, it is clear that ParaView since is for visualisation is a GUI based application. However, is there a way to do the “computations” without having to display anything on the screen? I say so since I would be running on a cluster, and since the main focus is MPI and not really the final visual rendering. I.e. using the tool from the CLI only. It goes without saying that if that’s even possible, then I’m interested in building it without any GUI support.
  2. Also, everything would be useless if not having a concrete test case with which using the application. For that, is there a public place where I could find some interesting tests to be used within ParaView? I am not interested in anything specific, only something that is large enough to be able to use MPI on a big scale.

Thanks in advance to anyone willing to give me help in this :slight_smile:

2

Yes, in a MPI context, people usually rely on pvbatch for that, a MPI-compatible pvpythin version.

For that, is there a public place where I could find some interesting tests to be used within ParaView? I am not interested in anything specific, only something that is large enough to be able to use MPI on a big scale.

There are many possible usecases for ParaView in distributed environnement, you may want to test I/O, computation performance, distributed rendering and so on.

A very simple workflow to test distributed processing would be to create a huge image data and then contour into it:

  • Wavelet, 1e? dimensions, Apply
  • Contour, Apply

Of course, you need to write this as a python script and then run it in pvbatch.

3

Thanks for your answer.

I’ll carefully read the docs on this.

I’m mostly interested in computation performance, since the tool I would like testing is supposed to provide accelerated MPI.

Only if you’d have suggestions on how to create that huge image data. I don’t think it can be any kind of data, can it?

4

Well, this is what I suggested, a way to create a huge image data procedurally.