SIMD parallel discrete-event dynamic system simulation

Abstract

We present experimental results for a mind-set to discrete-event dynamic system simulation. Rather than attempting to distribute the execution of each design experiment over a parallel single instruction multiple data (SIMD) machine's multiple processors, we simultaneously simulate a different experiment on each processor. We consider two experimental examples representing possible data communication, manufacturing or traffic networks. Each example's simulation results demonstrate the efficient combined implementation of the standard clock and ordinal optimization techniques versus parallel "brute force" simulation.