Abstract: For periodic hard real-time tasks running on homogeneous multi core processors, an energy-efficient approach based on dynamic voltage scaling (DVS) was designed. First, computation tasks are ordered by decreasing cycles and task mapping is arranged based on the principle of the shortest scheduling length for computation tasks. Then the computation task with the minimum commun- ication time is set as the last executed computation task while the order of other computation tasks keep unchanged for each processor core. The optimal execution order of computation tasks on each processor core is determined during execution time extension in the case of all computation tasks arranged with the highest frequency in the initial mapping. Experiments were conducted on several random task sets based on the power model of the Intel PXA270. Results show that the proposed approach can decrease energy of multi-core processors eff ectively.

Key words: Task graph, Energy-efficient, Real-time systems, Dynamic voltage scaling, Multi-core