... I think the bigger issue is why there is a PHDF5 file type, i.e. I don’t think cg_get_file_type should be returning CGIO_FILE_PHDF5 (also note that this is not documented). There isn’t a “parallel” HDF5 file format, it’s just an HDF5 file (which may have been created in parallel, and can be read in parallel). In the CGNS code, it for some reason gets the CGIO_FILE_PHDF5 file type if you built CGNS with —enable-parallel, but there is nothing inherently different in the CGNS file itself which makes it parallel.