File formats for distributed visualization

ParaView Support
1

Hello,

Using what file formats should simulation results be written in order to visualize simulation results on very large meshes using Paraview on a cluster? Among those file formats, is it possible to define blocks such that the “multi-block inspector” would be used to select which regions to visualize by ticking/unticking each corresponding box? Is this partition into blocks necessarily the one used to distribute over several processors?

Many thanks,

Sam

2

From what I found online in the documentation of paraview, I consider two file format for visualization in parallel: vtk (pvtu) and xdmf (.xmf) + hdf5. Would you have a recommendation or an alternative?

With xdmf + hdf5, I used the following structure to visualize results over several time steps

... (attribute, topology and geometry) ... ... ... ...

where I used the partition into several Region_k for two different reasons:

  • the multi-block inspector (that can be found under the View menu) can be used to select which regions to visualize by ticking/unticking their corresponding box
  • with the paraview server on N procs, Region_k will be distributed on the different processors with rank equal to k modulo N.

In general, we would be interested in visualizing selected regions of interest on distributed meshes such that the region of interest does not correspond to the partition onto the different processors. Is this a limitation of paraview or is there a workaround for doing so?

As far as I understand the vtk format, .pvtu files simply refer to several vtu that are distributed on the different processors but the part corresponding to each vtu cannot be selected in the multi-block inspector. Is this a limitation with vtk format or is there a workaround for doing so?

Thank you,

Sam