VTKHDF: clarification on partitions
The documentation for the VTKHDF file format for unstructured grids (here) talks about partitions, which seem to be mapped to MPI ranks. It would be great if the documentation could clarify some of the following points:
- Does the number of partitions at file creation influence/dictate the number of partitions I may/need to use during reading with ParaView?
- I assume the points do not have to be globally unique. Does that mean that, e.g., ParaView will not recognize cell connectivity across partition boundaries?
- Is there a downside to writing everything as a single partition (which should not be hard with HDF5, since it is possible to write to a single dataset in parallel)?