Algorithmic Skeletons for Parallelization of Embedded Real-Time Systems

Since parallel code is difficult to understand, it is a challenging job to develop it. It is even harder for code applied in embedded real-time systems. To facilitate parallelization of code running on these systems, we introduce a new algorithmic skeleton library. It is characterized by a deterministic execution behavior and is built in a timing-analyzable way. Our library offers algorithmic skeletons for parallel execution of tasks, parallel data processing and pipelined processing of data. When parallelizing a legacy single-core application, e. g. by applying a patternsupported parallelization approach, it may be employed to implement parallelization on code level. Furthermore, it can be utilized for enabling structured parallelism at development of a new real-time application for an embedded system.