Minimizing the cardinality of a real-time task set by automated task clustering

Abstract : The objective of this paper is first to properly define the notion of task clustering. This is the process of automatically mapping functionalities (blocks of code corresponding to a high-level feature) with real-time constraints to tasks (or threads). We aim at reducing the number of tasks functionalities are mapped to, while preserving the schedulability of the initial system. Second, our goal is to expose the complexity of the problem and to sketch methods we will propose for solving this problem. We consider independent tasks running on a single processor.