A Study On A Class Of Low Power Scheduling Algorithms In Heterogeneous Many-core Chip

With the sustained and rapid development of the semiconductor technology,the inside of many-core chip can be integrated with hundreds of processor cores.Providing hardware resources more parallelism,multi-core or many-core chip also makes its task scheduling and resource management become the key issue in the area of improving the computing efficiency and playing it fully.In addition,the "green computing" has become an important research direction in the field of information technology.However,it has become an unavoidable problem which is how to control and reduce their energy consumption,optimize and improve computing efficiency from the super computing system to desktop computer,and to a wide variety of intelligent computing terminal.What's more,improving the overall performance and reducing power consumption are both needed to be carefully weighed because of its many processor cores in many-core chip.This article focuses on how to improve the computing performance by reasonably schedule the application load and reduce the power consumption of many-core chip as much as possible.Directed Acyclic Graph(DAG)and Conditional Task Graph(CTG)can be used to model the different characteristics of the corresponding application.And this paper is aimed at studying the low power scheduling algorithms based on the DAG model and the CTG model in the heterogeneous many-core chip.This paper mainly conducts the following research:In the scheduling of heterogeneous multiprocessor based on DAG model,the task duplication strategy is to copy some specific tasks to multiple processor nodes,which can reduce the completion time of the multiprocessor systems program.Existing DAG low energy scheduling algorithms usually reduce the energy of the whole multiprocessor system by collaboratively considering the execution time and energy consumption of the application in the process of scheduling.Combining the task duplication strategy with the existing DAG scheduling algorithm,this paper proposes an energy aware DAG scheduling algorithm based on duplicated heterogeneous many-core chip,which regards heterogeneous many-core chip of grid structure as its system platform.The two indicators,the execution time of application and energy consumption,should be equally considered as the influence of the system performance in the process of scheduling.It can reach the aim of reducing the completion time of application and power consumption through introducing the task duplication strategy which reduces the delay and the power consumption caused by communication between tasks.Simulation experiments show that the proposed algorithm is effective to improve the performance of the application and reduce the power consumption of the system.In the study of the multiprocessor scheduling based on CTG model,the researchers mainly consider energy consumption as the optimal goal.This paper proposes a CTGoriented hierarchical scheduling algorithm of multiprocessor.The algorithm is divided into two stages: the first stage reduces communication overhead between tasks by task clustering,further load balance the processors and optimize the ta sk completion time on the basis of task clustering;Looking for the processor idle execution time in the process of the second stage is to find out the tasks before the idle time or after the idle time which match the condition of task stretching(extend the execution time of task),reduce the task execution energy by scale the task voltage in the case of guaranteeing application completion time,finally reach the goal of reducing the application completion time and lower the energy.The randomly generated CTG is used on simulation experiment and the results show that the proposed CTG low-power scheduling not only reduces the overall length of scheduling but also relatively lower the power consumption of the system.