Overview

Abstract. This chapter deals in a specific application domain with one of the main problems of component-based software development: the derivation of properties of a component system from properties of single components and rules for their interaction. Whereas properties of components can be analyzed and described by a software engineer responsible for component construction, many properties of an overall component system cannot be guaranteed in advance. This problem arises as a consequence of the late composition of components, for example by application engineers working for third parties. In order to overcome this problem in the area of measurement, signal processing, and control in embedded systems (M&C for short), components encapsulating signal processing algorithms, signal adaptation algorithms, or control flow mechanisms are modeled by means of a hierarchy of dataflow languages: synchronous (SDF), boolean controlled (BDF), and dynamic (DDF) dataflow. If an application engineer responsible for component assembly restricts himself to the utilization of SDF components, the component system is computationally analyzable. That is, it can be decided whether the system is deadlock-free or not, the amount of memory required can be determined, and a cyclic schedule of component instances can be computed. If the application engineer uses SDF and BDF components only, the component system is still deterministic. When DDF components are used, even determinism is lost.

The objective of this chapter is to describe a novel concept...