This paper describes two implementations of a semiconductor device simulator on two different architectures. The simulation process is based on the Monte Carlo method, a particle-in-cell algorithm. It simulates the motion of particles under the influence of electric fields, that have to be periodically updated with the Poisson equation solution. We present a parallelization using a bi-parallelism scheme, to avoid load imbalance that can be encountered with the domain decomposition. The parallelization of the Monte Carlo part gives optimal speed-up. However the poor efficiency of the implemented Poisson solution damaged this result.
展开▼