Closed rodionstepanov closed 10 months ago
Hi @rodionstepanov,
this is not (yet) possible. Full checkpoints would need storing the density distribution functions too, basically a full dump of VRAM, which means checkpoints would be gigantic in file size. And extension support would be rather complicated. Since simulation runtime is a only few hours at most, I decided not to implement checkpoints in the near future.
Regards, Moritz
Hi @ProjectPhysX ! Thanks for explanation. I agree the full checkpoints is too expensive but why don't take velocity field form stored data file and generate the density distribution functions as for an initial state?
Yes you can do that, use the .vtk output of one simulation as starting conditions for another. A .vtk loading/import function is not yet implemented but you can reverse-engineer the write_vtk() storing/export function.
It seems I can save current data with lbm.rho.write_device_to_vtk(); lbm.u.write_device_to_vtk(); but how to read it in order to continue simulation form the save data?