A cache-aware scheduling algorithm for embedded systems

Abstract

The paper presents a methodology for scheduling real time tasks in embedded systems where the task layout is known at design time and does not change at execution time (static systems) and where the cache miss costs are significant when compared to the normal execution time of the tasks. The scheduling model assumes a time driven dispatching of the application tasks which are ordered in a pre defined sequence. Building such a sequence in a way that is not only efficient but accounts for optimal cache sequencing is the aim of our method. The refinement of the schedule towards an optimal solution is done by simulated annealing techniques. The evaluation of the schedules is done by considering the effects of instruction caching when evaluating the computation time of the tasks.