摘要
在多处理器系统中, 寻求一个有效的并行任务设计安排使得整个执行时间达到最小是至关重要的环节。分割式遗传算法(PGA) 用来解决任务规划问题可以大大缩短规划时间, 但规划结果不甚理想。将分开逐个处理的思想与单、多点交叉概率分配、自适应变异概率设计相结合, 提出一种改进的分割式遗传算法。实验表明, 改进的PGA算法, 在进一步缩短规划时间的基础上, 提高了算法收敛速度和效率, 能够得到期望的更好的规划结果。
In the multiprocessor systems, an efficient scheduling of a parallel program onto the processors that minimizes the entire execution time is vital. This problem solved by Partitioned Genetic Algorithm (PGA) can dramatically decrease the time doing scheduling, but obtain non-ideal performances. Therefore, we propose a modified PGA, which integrates the concept of Divide-and-Conquer mechanism to partition the entire problem into subgroups and solve them individually ,with the idea to distribute the proportion of one-point crossover and multi-point crossover and to design the adaptive proportion of mutation, to overcome the drawback above. According to our experimental results, the modified PGA can not only decrease the time doing scheduling much more, but also improve the restraining speed and the efficiency of the algorithm, and obtain the anticipated performances, sometimes it is even much better.
出处
《计算机测量与控制》
CSCD
2005年第5期488-490,共3页
Computer Measurement &Control
基金
西安工业科技攻关计划资助项目(GG200368)
