A Research Of Multi-core Task Scheduling Algorithm Based On Adaptive Chaos

With the development of high-performance processor, heterogeneous multi-core processor becomes the main processor of modern processor recently. Only through taking the appropriate task scheduling, multi-core processor could perform its advantage of high performance. However, the task-scheduling problem of heterogeneous multi-core processor is a NP-complete problem. Most existing task-scheduling algorithms contain lots of problem, such as low efficiency in task scheduling, unreasonable task scheduling. Hence, it is necessary to design new task-scheduling strategy.In this paper, a new method that builds a DAG graph and task model to solve the multi-core task-scheduling problem, which is based on existing basic principle and strategy of task scheduling, was proposed. Ant Colony Optimization Algorithm and Genetic Algorithm to build were adopted to build the model. According to the simulation result, it shows that Genetic Algorithm does not perform well during later period, while the Ant Colony Optimization Algorithm performs poorly during the early period. To improve the performance of our algorithm, we combine the genetic algorithm and ant colony optimization algorithm. For our algorithm, we use the randomization and ergodicity of chaos theory to initialize the population. During the early period, we keep the optimal solution of every generation and optimize the result with chaos theory. In the later period, we leverage the positive feedback of ant colony operation and keep optimizing our result with chaos theory, which could improve the variety of solution and the global search ability, in case solutions constrict too soon.To validate the feasibility and the efficiency of our algorithm, we generate 3 sets of random tasks with random task generator TGFF. Then, we run genetic algorithm, ant colony optimization algorithm, and our new algorithm to evaluate, respectively. The simulation results illustrate that our algorithm performs better with high constriction and high efficiency in processing the multi-core task scheduling, which represents that our approach is meaningful for the development of task-scheduling problem.