lagrange

Jon and I were talking about how to parallelize the room-size computer. Interestingly our main debate seems to be an old-one. The Eulerian perspective is to divide the particles according to their locations in space, while the Lagrangian perspective divides particles into arbitrary groups regardless of position.

I found this article comparing the two methods to be very instructive. To cut to the chase, it basically says that the Eulerian approach scales better than the Lagrangian one for very large numbers of particles (N). Euler scales like N/P (P = number of processors), whereas Lagrangian scales like N over sqrt(P).