>> My question: is there a way, with the current code, to avoid that? >> For instance am I allowed to create an engine doing a loop over all the >> bodies and calling inside the loop, for each body, the engines >> MagnetEngine and GravityEngine. >> > What you suggest would work (some would call it a "hack" though...), > except that you need at least two engines : one for loops on bodies and > one for loops on interactions. > Not sure how much time it would save, I'm really wondering too. > You would need more, since typical scenario mixes loops over bodies and interactions like this:
(0. reset physical actions, increate time by dt etc.) 1. loop over bodies: create bounding boxes, collide them, create contacts. 2. loop over contacts: create interactionGeometry and interactionPhysics for them; use some material law to deduce forces on bodies in interaction. 3. loop over bodies: to apply other forces (gravity etc), get accelerations, integrate and displace. The module spherical-dem-simulator (which I never used): non-modular spherical DEM that could be benchmarked against the same scenario with regular engine-based architecture. Since Bruno said we are like 30% behind PFC for dem (which I would think is monolithic), I would be inclined to attribute those 30% to modularity. Regards, Vaclav _______________________________________________ Yade-users mailing list [email protected] https://lists.berlios.de/mailman/listinfo/yade-users
