Research On IFWA Task Scheduling Algorithm Based On Heterogeneous Chip Multi-processors

With the popularization and development of the Internet,there are more and more user data and more and more complex tasks.How to give full play to the performance of heterogeneous chip multi-processors is one of the hot issues in the field of computer architecture.The task scheduling algorithm of heterogeneous chip multi-processors directly affects the utilization of processor cores.so it is necessary to study an efficient,fast and stable task scheduling algorithm to improve the utilization of processors and give full play to the performance advantages of heterogeneous chip multi-processors.In this paper,the commonly used task models are analyzed,and a task scheduling model is proposed according to the characteristics of heterogeneous chip multi-processors.Then it makes an in-depth analysis and research on the basic principle of the fireworks algorithm,points out three shortcomings in the algorithm and puts forward some improvement measures.First,an initialization method based on fitness value and distance is proposed,in which the concept of fireworks distance is introduced,then 2N fireworks are initialized randomly,the calibrated fireworks with the best fitness value are found,and the distances between the rest of the fireworks and the calibrated fireworks are calculated,N fireworks with uniform distance distribution are selected as initialization fireworks to reduce the randomness of the initial fireworks population.Second,a mirror mapping strategy is proposed,aiming at the sparks beyond the feasible range,the fitness value of the fireworks generating the sparks is first calculated,and the sparks generated by the fireworks with good fitness values are abandoned directly and not mapped,so as to avoid the waste of resources caused by repeated calculations,only mirror mapping the sparks produced by fireworks with poor fitness value to enhance the local search ability of the algorithm.The third is to put forward the mutation-selection strategy,which integrates the idea of cross-mutation into the original selection strategy,and carries out the cross-mutation operation between a part of the fireworks selected by the roulette strategy and the excellent gene population to enhance the exchange of information between contemporary populations and offspring populations and accelerate the convergence speed of the algorithm.Finally,the fitness function which accords with the task scheduling problem is designed,the way of mapping between the fireworks algorithm and the task scheduling problem is proposed,and the improved fireworks algorithm-IFWA is successfully applied to the task scheduling problem.In this paper,IFWA algorithm,fireworks algorithm and genetic algorithm are simulated on Matlab platform.Comparative experiments are designed from three aspects: the number of tasks,the number of processor cores and the size of CCR parameters.The experimental results show that IFWA algorithm has a faster convergence speed in solving task scheduling problems,and the task scheduling sequence is shorter,which improves the utilization of processor cores.