Research On Dynamic Task Scheduling Strategy Based On Heterogeneous Multi-core Processors

With the urgent need for multi-core processor and related technologies of high performance computing in deepening awareness,The experts in abroad and native point out the merits of task scheduling algorithm paly a decisive role in capability of multi-core processor.Existing task scheduling algorithm are mostly based on static task on CMP,however,relatively few studies on heterogeneous multi-core dynamic task scheduling algorithm.Due to the dynamic task scheduling algorithm needs to collect the execution time of task,communication costs and data relationship of dependent tasks in real-time,so it will generate additional computational costs,but the efficiency of the scheduling algorithm is higher so that more dynamic task scheduling of multi-core processor become a hotspot in the field of task scheduling.On the basis of depth studying of heterogeneous multi-core scheduling algorithm,and fully considering the utilization of the communication cost,the load balancing of the processor's core and the communication costs of task,this paper put forward a dynamic task scheduling algorithm about heterogeneous multi-core processors.Firstly,the initial task of the DAG is converted to the task list which contains status bits;then calculate the earliest completion time of pending task in the task list,choose the task scheduling node of minimum earliest completion and calculate whether can further reduce time of task completion by task duplication technique.If duplication technique can further reduce the completion time of task,the task is to copy.Otherwise not,which can prevent the generation of redundant nodes.In addition,the task scheduling algorithm set load threshold for each processor to solve the overloading problem of some processors to achieve the load balancing of multi-core processor.In order to validate the dynamic scheduling algorithm has a better scheduling performance on heterogeneous multi-core platform,the paper do a series of reasonable comparative experiments on Simics simulation platform.The results show that,dynamic task scheduling algorithm proposed in this paper can improve the degree of implementation of tasks on parallel processors,reduce task completion time and ensure load balancing between each processing core,thereby effectively improve scheduling efficiency of the processor.