Research On Task Scheduling Optimization Of Heterogeneous Multi-core Processors With Energy Constraints

Heterogeneous multi-core processors are widely used in smart phones,automated robots,wearable devices and other mobile embedded electronic products because of their excellent computing performance and parallel task processing ability.However,due to the slow development of battery technology,mobile devices need to meet certain energy consumption limits while improving performance.The commonly used DVFS technology can reduce energy consumption,but it also brings performance degradation.Therefore,it is necessary to further optimize the performance and energy consumption of heterogeneous multi-core processors on the basis of DVFS through reasonable task scheduling methods.The research on this problem mainly focuses on how to minimize the application completion time,that is,to minimize the overall scheduling length of the application under certain energy consumption constraint,so as to improve the system performance.At present,the commonly used method is scheduling based on task energy consumption weight,but there are still some shortcomings in energy consumption preallocation strategy,task priority determination method,local optimal characteristics of scheduling and so on.In view of the above shortcomings,this paper proposes a method based on task hierarchy and scheduling post-processing.This method mainly includes four parts.Firstly,aiming at the problem that the existing methods consider too few factors in task energy pre-allocation,this paper proposes an energy pre-allocation strategy considering task hierarchical structure,which improves the rationality of energy pre-allocation and shortens the application scheduling length;secondly,based on the task energy preallocation strategy and traversal method,a new task scheduling algorithm is proposed;thirdly,a dynamic priority adjustment mechanism is proposed for the traditional way of task priority determination,which makes the determination of task priority more conducive to shorten the scheduling length of the application;fourthly,according to the local optimal characteristics of the scheduling process,a task frequency re-adjustment mechanism is proposed,that is,after the initial task scheduling,the execution frequency of the task is re-adjusted to reduce the negative impact of the local optimal characteristics,so as to further shorten the adjustment length of the application.In this paper,based on real parallel applications and random applications,by adjusting the energy consumption constraint or number of tasks,a large number of comparative experiments are carried out to verify the effectiveness of the method.Experimental results show that,compared with other existing methods,the proposed method can obtain shorter scheduling length under the same energy consumption constraint,and the scheduling length is reduced by about 6% ? 15% compared with the latest existing method.