Research On Energy-Efficient Mixed Task Scheduling Algorithm On Heterogeneous Multi-Core SoC

With the popular application of heterogeneous multi-core architecture in embedded real-time systems,the energy consumption problem is increasingly prominent.To solve this problem while improving the efficiency of task execution and ensuring that tasks do not exceed their deadlines,energy-efficient task scheduling algorithms have caused serious concern of academia and industry.The main contributions of this thesis are as follow:1.An energy-efficient scheduling algorithm based on convex optimization theory for mixed task is proposed in details.Most of the existing algorithms depend on simple task models,so can not be applied in complex scenarios.The algorithm proposed considers a mixed task model of periodic and aperiodic tasks,which is closer to the realworld scenarios.The algorithm models the scheduling problem as a programming problem with the objective of minimizing the total energy consumption,and then relaxes it into a convex optimization problem for being solved.Experiments show that the energy consumption of the algorithm is lower than that of the heuristic by 48.7%～69.6%,and the execution time is lower than that of the similar.The feasibility of the scheduling scheme and the response to the aperiodic task are better than the comparison.2.On the basis of the above algorithm,an improved energy-efficient scheduling algorithm for mixed task is proposed in details.The algorithm assigns task to corresponding processor according to the maximum task density difference of periodic tasks and the maximum and worst-case execution time difference of aperiodic tasks.Therefore,it avoids long execution time and difficult implementation like the above algorithm.The idle time generated by the early completion of the task further reduces the frequency of the processor,thereby reducing the total energy consumption.Experiments show that the energy consumption of the improved algorithm is 25.3%～38.7% lower than that of the heuristic.It takes much less time to execute than the above algorithm and no tasks exceed their deadlines.3.Most of the existing task scheduling research based on heterogeneous multi-core platforms only depend on the ideal energy-model.The application scheme of the proposed scheduling algorithm on the Sitara series AM5728 of TI is considered in this thesis.And then,based on the actual system,a serious of experiments are designed to verify the algorithm.Experiments show that the two algorithms proposed in this thesis perform well in terms of energy saving,feasibility of scheduling scheme,and response to aperiodic tasks.The improved algorithm determines that the scheduling scheme has a shorter execution time and is easy to implement.