Introducing a Memory Efficient Execution Model in a Tool-Suite for Real-Time Systems

Abstract: This thesis shows how development of embedded real-time systems can be made more efficient by introduction of an memory efficient execution model in a commercial development suite. To identify the need of additional support for execution models in development tools, the thesis investigate by a series of interviews, the common requirements in development of industrial embedded real-time systems. The results indicate that there exist functionality in industrial systems that could be more efficiently implemented in other execution models than the currently supported ones. The thesis then presents how use of multiple execution models (hybrid scheduling) can reduce processor utilization in real-world applications. Furthermore, the thesis presents an integration of an memory efficient execution model in an industrially used real-time operating system. In addition, the thesis describes an efficient technique to analyze memory consumptions of functionality executing under the introduced execution model.Embedded computers play an important role in peoples everyday life. Nowadays, we can find computers in product such as microwave ovens, washing machines, DVD players, cellular phones and cars, to mention a few examples. For example, a modern car may have more than 70 embedded control units handling functionality such as airbags, anti-lock braking, traction control etc. In addition, there is a clear trend indicating that the amount of computer controlled functionality in products will continue to increase. Many of today’s embedded systems are resource constrained and the software for them is developed for a few execution models. Even though researchers has proposed a numerous number of different execution models for embedded real-time systems, in practice however, only a few of the proposed execution models are supported in industrial development tools. This implies that developers often force fit functionality to be executed under these models, resulting in poor resource utilization and increasing complexity in software.