This configuration option does more damage than good because computations with complex boundary interactions tend to fail. Let's bite the bullet and simply always use boundary d_ijs for the computation of the maximal time step size.
Incidentally, skipping the lookup into the boundary map via count() improves performance quite a bit and allows us to refactor the "BoundaryMap" into a vector.
This configuration option does more damage than good because computations with complex boundary interactions tend to fail. Let's bite the bullet and simply always use boundary d_ijs for the computation of the maximal time step size.
Incidentally, skipping the lookup into the boundary map via
count()improves performance quite a bit and allows us to refactor the "BoundaryMap" into a vector.