I’m generating VTKHDF unstructured grid datasets in parallel consisting of multiple parts. Gemini tells me that if I include a vtkGlobalPointIds array in the PointData group, Paraview will use it to better stitch together the grids from the individual parts. Is this true? I can certainly see that point array in PV, but I can’t tell if its being used for that purpose – I wasn’t seeing any display artifacts before that might suggest it.
Similarly, gemini suggested as an alternative, including ghost cells in my output and writing a vtkGhostType array in the CellData group.
Neither of these things is mentioned int the VTKHDF specification and I’m unfamiliar with VTK in general.
Are there other optional things that can be added to the VTKHDF file for unstructured grids?
Not specifically, GlobalPointsIds is just data that can be used by certains filters to do certain computation. GhostCells are way more important though. It has nothing to do with VTKHDF specifically.
including ghost cells in my output and writing a vtkGhostType array in the CellData group.
Indeed, critical to get correct computation on distributed data, although you can compute them using GhostCells filter.
Neither of these things is mentioned int the VTKHDF specification
Because they have nothing to do with VTKHDF and all to do with VTK.
In the context of VTKHDF, ghost and global ids are just point and cell data.
If I understand you correctly, Paraview filters will automatically look for point and cell arrays like vtkGlobalPointIds and vtkGhostType and use them when they can? Gemini said that the PointData group needs to have an attribute GlobalIds that specifies the point array to use as vtkGlobalPointIds (chatgpt disagrees). If true, that would seem to be a VTKHDF-specific detail. But I’m guessing that nugget of wisdom is false.