Particle-Mesh Methods on the Connection Machine

Abstract

We describe an efficient Particle-Mesh algorithm for the Connection Machine CM-5. Our particular method parallelizes well and the computation time per time step decreases as the particles become more clustered. We achieve floating-point computation rates of 4--5 MFlops/sec/processing node and total operations (the sum of floating-point and integer arithmetic plus communications) of 5--10 MOps/sec/processing node. The rates scale almost linearly from 32 to 256 processors. Although some of what we discuss is specific to the CM-5, many aspects (e.g., the computation of the force on a mesh) are generic to all implementations, and other aspects (e.g., the algorithm for assignment of the density to the mesh) are useful on any parallel computer.