Research On Optimization Strategy Of Genetic Algorithm Implementation Of Multi-core Processor Based Tasks

Along with the development of modern science, technology, military, medicine and life in all aspects are required to improve the performance of the computer. The traditional way to improve the frequency of the processor to improve the performance of the computer has reached the limit, and the electronic production process level also entered the bottleneck period. In this case, the multi core processor came into being, it can improve the performance of computer by increasing the number of processing cores.The structure of multi core processor is more complex than that of single core processor, and the resource management is also facing a more severe test. When the task is executed on multi core platform, how to make full use of multi core hardware resources that the task can be executed quickly and correctly, becomes a research hotspot of modern multi core processor. The task on multi core hardware platform are often expressed by D AG diagram, the deficiency of this is:the meaning of nodes in the model is not clear; the reasons for the impact of the communication between the nodes on the task execution are not clear, and so on.In the light of the shortage of traditional DAG diagram in describing the execution of the task, this paper designs and implements the task execution model, and makes the following improvements: redefine the meaning of the nodes in the model; the amount of communication and task execution time are represented by the scheduling times of the system; the impact of communication o n task execution can be expressed by the length of time; in conjunction with the execution time of the task traffic between nodes to consider load balancing between processing cores, this will makes the implementation of the mandate closer to real. Then us e the degree of load balancing between core processors as a criterion, Task Execution Model perform better than DAG diagram in the degree of load balancing between core processors. After tasks on multi core hardware platform is expressed by the Task Execution Model, this paper uses genetic algorithm to solve the scheduling sequence of tasks on multi platform, all the steps of genetic algorithm are redesigned as follows: individual encoding, initial population generation, fitness function, genetic operator design(selection strategy, cross operation, cross operation) and the algorithm ends condition. Finally, genetic algorithm and the task scheduling algorithm of multi-core parallel system is compared by experiments. The experimental results use time difference between multi-core and the total time of the task execution as evaluation criteria, proved that the genetic algorithm designed in this paper can perform better in the process of the load balancing, which makes the task execution can be executed quickly.