WIP: Parallel unskip
In zoom simulations, a single top level cell can have a large fraction of all the cells. Therefore a parallelization of engine_unskip
over the top level cells might not be a good approach. Here I am adding a parameter that decide the level of the parallelization (Scheduler:unskip_depth
0 means top level [as before] and a higher value will push it down in the tree).
Currently, in this branch, we are not faster, therefore it should not be merged. Matthieu is working on the cell_unskip
and I will update this branch once his work is done.
Here we are stuck due to the gravity activation. A possible improvement would be to use a flag in order to skip early the recursion in cell_activate_subcell_grav_tasks
.