Parallel and Distributed Three-Dimensional Monte Carlo

We present a parallel Monte Carlo algorithm for the simulation of semiconductor devices in three dimensions. The physical behavior of the system is governed by the Boltzmann Transport Equation. In the absence of direct interactions among charge carriers, the samplings of the statistical space are independent. This results in a potential decoupling of computational tasks that reduces the number of required communications in simulation. A Monte Carlo algorithm, with both static and dynamic load balancing strategies, is shown for the Connection Machine CM-5 and a network of workstations. We show that both load balancing strategies scale well with problem size, and present results on relative speedups between them. A randomized analysis of performance is presented and physical simulation results are shown.