Abstract
This paper describes the creation of an intelligent software tool that performs automatic parallelization of dynamic system simulations for excecution on arbitrary message-passing multi- computing configurations. The tool greatly decreases the amount of time needed to obtain a viable multiple processor implementation of large-scale simulations and can be expanded to include the latest methodologies that exploit functional and data parallelism present within the simulation model and algorithm. Such a unifying tool is designed to increase the likelihood that a parallel realization of a simulation problem can be obtained which has an acceptable level of performance. To illustrate the feasibility of developing such a tool, this paper describes the prototype implementation of the Automated Partitioning and Mapping Engine (APME) and demonstrates its effectiveness when applied to a large-scale simulation executed on a number of multicomputer systems and topological configurations.
Get full access to this article
View all access options for this article.
