Energy-Aware Task Scheduling with Precedence and Deadline Constraints on MPSoCs

Energy reduction is a major issue in designing embedded systems. We investigate the problem of minimizing the total processor energy consumption of a set of nonpreemptible tasks with precedence constraints and individual deadlines that are executed on an MPSoC (MultiProcessor System on Chip) without shared memory, and propose a unified approach under two power models, namely the dynamic power model and the total power model. Our approach employs a novel priority scheme for task assignment and uses NLP (NonLinear Programming) to assign an optimal execution speed to each task. We have implemented our approach and compared it with two state-of-the-art energy-aware task scheduling approaches, namely LL-ES-GREEDY and EES by using a set of synthetic and real-world benchmarks. Experimental results show that the maximum improvement, the average improvement and the minimum improvement of our approach over the LL-ES-GREEDY approach are 42.44%, 30.46% and 9.46%, respectively. The maximum improvement, the average improvement and the minimum improvement of our approach over the EES approach are 75.98%, 39.74% and 7.08%, respectively.