> Do you think recalculating a bounding tree after every simulation step
> will be faster in the end?

        I don't know how collision detection is performed in MegaPOV for now: does
it account for all particles at once ?  (i.e. O(N^2) time complexity ?)

        From my experience in biomolecular simulations, there is in principle no
need to recalculate the interaction list (or bounding tree) at every step
of the simulation.  Simple heuristics can be used to recalculate it when
the particule moves away by more than a certain distance from the position
at which the bounding tree was calculated.  Of course the efficiency then
depends on the velocity of the fastest particle at a given time step, but
typically the tree could be updated only every 5-10 steps.  Speed up is
considerable with respect to O(N^2) anyway (especially for short "cutoff"
distance when calculating the interaction list).

        - NC

Post a reply to this message