This PR implements the option to generate binary HDF files for the solutions of the dmd/parametric_heat_conduction.cpp example. Previously, CSV files were the only option. CSV is still used for some small amounts of data describing the simulations, but HDF can be used for solution data. The advantage of the HDF option is that the data is stored in full precision more efficiently than CSV, and there is only 1 HDF file per simulation as opposed to one CSV file per timestep. For example, heat_conduction_hdf.sh runs several times faster than heat_conduction_csv.sh for generating training data. The Database classes are enhanced to support the new features.
This PR implements the option to generate binary HDF files for the solutions of the
dmd/parametric_heat_conduction.cpp
example. Previously, CSV files were the only option. CSV is still used for some small amounts of data describing the simulations, but HDF can be used for solution data. The advantage of the HDF option is that the data is stored in full precision more efficiently than CSV, and there is only 1 HDF file per simulation as opposed to one CSV file per timestep. For example,heat_conduction_hdf.sh
runs several times faster thanheat_conduction_csv.sh
for generating training data. TheDatabase
classes are enhanced to support the new features.