Domain decomposition strategies for the two-dimensional Wigner Monte Carlo Method

A domain decomposition approach for the parallelization of the Wigner Monte Carlo method allows the huge memory requirements to be distributed amongst many computational units, thereby making large multidimensional simulations feasible. Two domain decomposition techniques—a uniform slab and uniform block decomposition—are compared and the design and implementation of the block decomposition approach, using the message passing interface, is discussed. The parallel performance of the two approaches is evaluated by simulating a representative physical problem. Our results show that the presumably inferior slab decomposition method is in fact superior to the block decomposition approach, due to the additional overhead incurred by the block decomposition method to set up its communication layer.