Open mobernabeu opened 7 years ago
http://vis.lbl.gov/Research/H5hut/ perhaps worth a look.
Thanks for the pointer, @uschille. HDF5 seems like the sensible way to go.
We started developing a H5MD-based solution in #560 and #561. Do you have any views on H5hut vs H5MD?
I don't have any hands on experience, but H5hut seems a bit more established (National Lab software tends to be reasonably well maintained), comes with mesh support, and can be read by Paraview. Not sure whether it's worthwhile switching though.
RBCs are currently written to disk in VTK's VTP ASCII format. For the RBC mesh made of 1620 facets, this is nearly 100K in size when force magnitudes are also included. For long simulations with dense suspensions, one may end up writing O(1K)-O(10K) files, which quickly takes the output size to the GB region.
Switch to VTK's VTP binary format and consider how difficult it would be to use compression without going through the VTK API for writing (#578).