Expressing intratask parallelism in discrete event simulation models

Abstract

For parallel and distributed simulation, a model is usually defined to consist of independent tasks which synchronize by communicating time-stamped events/messages. In this paper, we define a discrete event simulation modeling paradigm which supports explicit expression of intratask parallelism, i.e., parallelism within a task. In this paradigm a task is defined by a set of rules whose actions are triggered based on multiple simultaneous event occurrences. New ways of expressing parallelism in a model emerges because (1) actions of multiple rules can be executed in parallel and (2) an action of a rule can be defined by data parallelism on parameter values of simultaneous event occurrences. The usefulness of the constructs of our paradigm is illustrated by an example.<>