Research On Task Scheduling And Memory Management Based On Multi-Core Processor

With the development of hardware technology,the multi-core processors get increasingly used in our daily life,scientific computing and other fields.The performance of the multi-core processors remarkably outperforms the traditional single-core processors,and they have the superiority of lower power consumption.To give full play to advantages of the multi-core processors as hardware,appropriate software technology must be established to support.Task scheduling is one of the key points that affect the performance of the multi-core processors.To make best of each core of the multi-core processor,an efficient multi-core task scheduling algorithm is indispensable.In addition,reducing memory access overhead in the multi-core environment is of great help to improve the performance of the whole processor system.This paper focuses on the task scheduling and memory management technology in the multi-core environment.Multi-core task scheduling has been proved to be a NP-hard problem,and it is solved basically by heuristic algorithms or genetic algorithms at present.The dynamic task scheduling algorithm of multi-core processor is studied in this paper.Among the existing algorithms,TA algorithm has a higher success rate in task scheduling,but it overlooked the core utilization and this leads to long idling of certain cores and uneven utilization between the different cores.To solve this problem,improvement to the previous algorithm is described in this paper with multi-attribute decision-making idea in the decision theory.In task scheduling,dynamic information and parameters of active cores are taken into consideration and the core with the highest comprehensive evaluation rate shall be picked for the task scheduling.Experimental results show that the improved algorithm has higher scheduling success ratio,core utilization ratio and balanced utilization ratio between cores as compared to the TA algorithm and LBTA optimization algorithm.In the multi-core environment,as a public resource,the memory will be competed for more intensely with the increase of cores and tasks.The existing memory management models of Ptmalloc,Tcmalloc,Jemalloc and etc.will go worse and the memory access efficiency will decrease if too many tasks are to be handled.Besides,access to the public resource requires corresponding synchronization mechanism in which the "starvation" phenomenon may occur if an individual task frequently competes for certain resources and other tasks are reduced to waiting and making no response.In this case,high parallelism of the multi-core is not used well.An efficient and balanced memory management model is proposed in this paper.The model adopted the hierarchical memory management,with a consideration of the private non-locking idea,and combined synchronous and mutual exclusion algorithm to improve the memory release efficiency for memory being requested and guarantee the balanced use of the cores and memory.Finally,a comparison is made between the proposed model,Ptmalloc,FPA memory management model of the experimental platform.The result shows that the new model is superior in terms of the memory access efficiency and fairness in competition for resource.