Starting from a checkpoint with different adaptivity parameters

[david-montiel-t]

There may be issues when restarting a simulation with different adaptive criteria than the original simulation.

For example, consider a simulation with minimum and maximum refinement levels set to 5 and 7, respectively. The simulation is run starting from time t=0 for Ns time steps and it outputs a the last checkpoint at time step Nc. If we restart the simulation from checkpoint corresponding to Nc but for the new simulation modify the minimum and maximum refinement levels to 4 and 6, respectively, the mesh in the region of the system corresponding to level 7 will not unadapt and stop evolving.

[landinjm]

This has to do with the refinement criterion logic & perhaps the order of loading restarts and calling for AMR in init.cc. Should be an easy fix.

phaseField/src/matrixfree/refine.cc

Lines 152 to 160 in 5f4829a

	//limit the maximal and minimal refinement depth of the mesh
	unsigned int current_level = t_cell->level();
	if ( (mark_refine && current_level < userInputs.max_refinement_level) ){
	cell->set_refine_flag();
	}
	else if (!mark_refine && current_level > userInputs.min_refinement_level) {
	cell->set_coarsen_flag();
	}

Keep in mind that we should warn the user that this is happening.