Sagar —
Inactivation of structures is primarily handled by zeroing out forces and velocities that are associated with those structures; I think that in all cases, the zeroed-out force is still spread onto the grid, and the velocity is still interpolated back to the structure before it is zeroed out. So there winds up being unnecessary computational work that is performed for the inactive structures. If a large fraction of the structures in a model are inactive, then that will make the computation much more expensive than it needs to be.
If this is a major bottleneck for you, and you are interested in working to improve support in the library for this use case, we’d be happy to try to come up with a plan.
— Boyce