Efficient Memory Utilization in Resource Constrained Real-Time Systems

This thesis presents design and run-time techniques for efficient memory utilization in embedded real-time systems. The proposed techniques give developers means to reduce the memory consumption in the systems. Altogether, this gives possibilities to increases the added value of industrial systems, in the sense that more features can be fitted on existing hardware. The thesis begins by presenting the results of a series of interviews concerning common requirements in development of embedded real-time systems. Based on these results, the thesis presents a novel component model for development of resource constrained real-time systems. The model supports efficient memory usage through stack sharing and is formal enough to enable predictability of the resulting stack usage. To provide run-time support for stack sharing, the thesis presents an integration of the stack sharing strategy in an operating system for the component model. To determine the resulting memory usage under stack sharing, a novel analysis method is presented. In an evaluation, the thesis show that the analysis method is both fast and that it gives tight bounds on the resulting stack usage, which makes it suitable for industrial use. The thesis ends with a presentation showing the integration of the proposed analysis technique in an integrated development environment. The proposed techniques have been integrated in the commercial tool Rubus-ICE from Arcticus Systems. The techniques will be available for developers in the upcoming release of Rubus-ICE.