Paraview Visualization

An embarrassingly parallel part surface extraction script using a MPI enabled build of pvpython follows. Download extractSurface.py and run it with the command:

mpirun -np numProcesses pvpython ./extractSurface.py /path/to/input/dir/someMesh.pvtu /path/to/output/dir

Assuming all goes well a pvtp file and sub-directories containing vtp files will be created in the /path/to/output/dir directory. It is critical that pvpython have MPI support. See the CCI wiki for info on the ParaView install.

Follow the CCI wiki instructions to visualize the *vtp files in Parallel on the CCI.

The script takes a .pvtu and its corresponding .vtu, computes the surface of each .vtu and produces a .pvtp and .vtp file. These files are no longer "unstructured meshes" but poly data files. They are a representation of the surface of each part, contained all the point data from the existing .vtu files. Cell data is lost in this format. By visualizing only the surfaces of each part, the individual files are reduced in size by between 50%-90%.

A loose rule of thumb is 1 second per 10 MB of vtu files (ascii file) per part to convert, so for a 16384 part mesh on 128 cores with 50 MB parts, 5 seconds per part * 128 parts per core ~ around 10 minutes should be sufficient.