Research On Task Allocation Algorithm With Energy-efficient Heterogeneous Multi-core Under Shared Resource Constraints

With the rapid development of computer technology,embedded devices are used more and more widely.In order to meet the processing demand of embedded devices for different tasks,heterogeneous multi-core processors are favored by the market.With the increase in computing power,the power consumption of the device is also increasing,which will not only reduce the operating hours of embedded devices,but also generate excessive heat,resulting in a decline in the user experience.Therefore,the low power consumption of heterogeneous multi-core embedded devices has become a current research hotspot.The focus of this thesis is to reduce the energy consumption of the system while ensuring real-time performance by scheduling tasks that access shared resources in a heterogeneous multi-core platform.Task partitioning and energy-efficiency scheduling are the two stages of energyefficiency task scheduling.In the partitioning stage,tasks need to be partitioned to the corresponding processor according to certain rules.The energy-efficiency scheduling stage reduces the energy consumption of processors and devices according to energy-efficiency strategies,including Dynamic Voltage and Frequency Scaling(DVFS)and Dynamic Power Management(DPM).In the task partitioning stage,a task partitioning algorithm based on resource similarity-based and energy density perception is proposed.It aims to reduce the waiting time when tasks request shared resources,and to ensure the schedulable rate while reducing system energy consumption.The algorithm introduces energy density difference,sorts the energy density difference of tasks,and partitions tasks in order.As far as possible,tasks with similar access to shared resources are partitioned to the same core to reduce resource access conflict time.Taking into account the heterogeneous characteristics of processors,the algorithm selects the processor with the most suitable energy density,thereby reducing system energy consumption.The schedulable rate of this algorithm is 2%?8% higher than that of existing algorithms;the energy consumption of this algorithm is 8%?12.5% lower than similar algorithms.In the single-core task scheduling stage,a single-core scheduling algorithm that integrates DVFS and DPM is proposed.The algorithm takes into account the energy consumption generated by the processors and the devices,and models the minimum energy consumption problem.By solving the mixed integer linear programming problem,the optimal start time and operating frequency of each task are obtained.In the process of solving problem,the device idle time model is established,and the virtual task perfect model is constructed.At the same time,the related variables are relaxed,and the constraint conditions are judged when the variables are rounded to select a more appropriate frequency.The final result can be obtained faster.The energy consumption of the algorithm is 2%?10% lower than that of the existing algorithm;the device sleep time is 9%?30% longer than the existing algorithm and works in a low-power state for longer.