Particle swarm algorithm based task scheduling for many-core systems
Lu Junliang, Wei Hu, Huan Shen, Yaxin Li, Jing Liu · 2017
Task scheduling is one of the key factors that determine the performance of the computer systems. When there are more cores on a single chip, how to schedule the multiple tasks to these cores is still important and also a challenge. In order to optimize the efficiency of task scheduling of many-core computer systems, this paper proposes a new task scheduling algorithm based on particle swarm ant colony algorithm. In this algorithm, the fitness function is chosen according to the system model with many cores and multiple tasks. This algorithm uses B-level scheduling method in the initial stage to generate the initial pheromone distribution. In the iteration process, the improved way is used to update pheromone to adjust operator. At the same time, the crossover and mutation strategy of genetic algorithm is used for adaptive crossover and mutation of local optimum and global optimum particles, which makes the location of particles be able to change and update according to the fitness value of particles. The experimental results show that the improved algorithm is superior to the genetic algorithm in the performance of task scheduling.