Parallel processing for system simulation

In this thesis, a relaxation algorithm is proposed for solving large scale system simulation problems in parallel. This algorithm is composed of both a time-step parallel algorithm and a component-wise parallel algorithm and fully exploits the interconnected nature of the system which is characterized by a component connection model. Furthermore, the possibility of this algorithm being implemented with the structures suggested by Pottle [6] and Van Ness [2] is discussed. In chapter 2, the component connection model of a dynamical system and the classical (sequential) relaxation algorithm are reviewed. In chapter 3, we describe not only the time-step parallel algorithm and the component is parallel algorithm, but also the combination of these two algorithms to formulate a "highly parallel" system simulation algorithm. Chapter 4 illustrates a number of examples in which the possible trade offs between efficiency, speed-up ratio, and waiting time are analyzed. Finally, some concluding remarks are made in chapter 5.