Hello Yu Liu,
CarpetRegrid2 only sets two levels on the grid structure.
INFO (Carpet): Grid structure (superregions, coordinates): [0][0][0] exterior: [-6,-6,-6] : [28,28,28] : [2,2,2] [1][0][0] exterior: [-3,-3,-3] : [26,26,26] : [1,1,1]
Why the course level evolves in time every 512 steps.
The reason lies in the Carpet::max_refinement_levels = 10 that you asked about below. This setting says that, during the course of the simulation, there may be up to 10 levels of mesh refinement, even if not all of them are active at the beginning. Thus for 10 levels and a factor of 2 between each level, the finest level steps every single iteration (1 = 2^0), the 2nd finest every 2nd iteration (2 = 2^1), then the 3rd finest every 4th iteration (4 = 2^2), until the 10th finest (ie the coarsest) level steps every 2^9 = 512 steps.
If fewer levels are actually active (as in this simulation) then the steps corresponding only to non-existing levels never execute.
Iteration Time | *me_per_hour | *ROBASE::rho | | maximum
0 0.000 | 0.0000000 | 0.0012800 512 0.500 | 2.545256e+03 | 0.0012770I noticed the parameter Carpet::max_refinement_levels = 10, but it does not seem to affect the grid structure.
This parameter declares the maximum possible number of refinement levels that can exist, not the actual refinement levels created. Those are controlled by the regridding thorn CarpetRegrid2 in its num_levels parameters.
I think this is important for us to set I/O and CarpetRegrid2::regrid_every.
Correct. Though regrid_every is also usually driven by physical time scales. Eg if you are tracking an object then regrid_every must be short enough that, given a typical speed, the object has not moved out of the refined box that is supposed to track it by the time the regrid happens and the box location is adjusted. Note that regridding more often then every coarse step is perfectly fine.
Yours, Roland