Solving task scheduling problem in multi-processors with genetic algorithm and task duplication

Parallel arithmetic are methods for processing in distributed and multi processors environments. The purpose of parallel arithmetic is to accelerate executing a group of tasks, dividing applications to sub-tasks and executing them at the same time. In this paper, we propose a genetic based technique for solving task scheduling in multi-processor systems. In some cases, the cost to execute a task becomes more than retrieving the information of task from one processor to another. To address this property we use a thought-out task duplication policy to decrease the overall computation time. Because each task can duplicate more than once, the length of chromosomes in the proposed method will change dynamically. Furthermore, a simple and efficient strategy is proposed for task priority assignment. Experimental results confirm the effectiveness of our proposed method in seven benchmark problems in comparison with previous works.