Energy-Performance Trade-Off for Multi-core Embedded Systems

This paper focuses on scheduling periodic tasks on systems with discrete frequencies to achieve energy consumption minimization without sacrificing performance. Based on regulated directed a cyclic graphs, an approach is proposed to solve the problem. An algorithm called list scheduling-based task scheduling (LSTS) algorithm is designed to complete task mapping and frequency selection. We conduct experiments with a set of task sets on the power model of the AMD Mobile Athlon4 processor. Experimental results show that this approach can achieve great energy savings compared with the list scheduling algorithm.