Research On Scheduling For Heterogeneous Multi-Core Processors

In recent years,the rise of artificial intelligence has led to a significant increase in the processor computing power requirements of many applications,and existing homogeneous multi-core processors are difficult to meet the growing computing power requirements of large applications.On a homogeneous multi-core processor,the execution cost of the same task on each processor core is exactly the same,and the heterogeneous multi-core processor can perform different processing for the characteristics of the task,and dispatch different types of tasks to the processor that adapts to the task,which can greatly reduce the completion time of the entire task,while also having lower energy consumption.Therefore,more experts and scholars have turned their attention to heterogeneous multi-core processors,and the scheduling algorithm on heterogeneous multi-core has become a hot research topic.Based on the existing static heuristic scheduling algorithm,in this thesis,a bi-objective and muti-strategy scheduling algorithm is proposed.The scheduling algorithm is divided into a priority calculation phase and a task assignment phase.In the priority calculation phase,the algorithm overcomes the single priority calculation standard,increases the weight of the task sending and receiving data,and enables the task with large communication data to be preferentially scheduled.In the task allocation phase,a multi-policy assignment method is used.One is to combine the task replication technology with the interval insertion technique to calculate the earliest completion time,and the other is to assign the current node and the precursor node with the longest communication time to the same processor core.The earliest completion time is calculated.The third is to allocate the current node and the child node with the longest communication time to its successor node to the same processor core to calculate the earliest completion time.The most utilized scheme for the idle time period is selected as the criterion for task assignment.Finally,the DVFS technology is used to optimize the energy consumption of the scheduling results,so that the task can run at a lower energy consumption without increasing the makespan.In order to verify the performance of the algorithm,two groups of experiments are conducted in this paper.In the first group of experiments,the effects of CCR and the number of tasks on the scheduling results are studied by randomly generating a large number of DAG task maps.The second group of experiments is to schedule the matrix LU decomposition,fast Fourier transform,2D-Wave programs.Finally,the experimental results are analyzed.The experimental results show that the proposed algorithm can effectively improve the scheduling efficiency of multi-core tasks and has a better prospect.