基于主从式并行遗传算法的S盒优化算法  

Optimum algorithm of S-boxes based on master-slave parallel genetic algorithm

作　　者：殷新春[1] 仇亮[1] 

机构地区：[1]扬州大学信息工程学院

出　　处：《计算机工程与应用》2008年第24期112-114,117,共4页Computer Engineering and Applications

基　　金：国家自然科学基金No.60473012;江苏省六大人才高峰(No.06-E-025)~~

摘　　要：针对传统遗传算法优化高阶S盒中适应度评价值计算量相对大的问题,利用MPI+C环境结合主从式并行遗传算法框架来改进传统串行遗传算法,在MPP处理机——"深腾-1800"(联想)上进行测试,实验结果表明,运用此并行算法优化6阶、8阶的双射S盒,只在优化8阶的情况下能得到更好的加速比,从而验证了主从模式在解决适应度评价值计算量大的问题时,可达到接近线性的加速比。The calculation time of fitness function evaluation in the optimization of S-boxes based on genetic algorithm was much great.We utilized master-slave parallel genetic algorithm with MPI environment binding for C to improve the sequential genetic algorithm and it was implemented on the Massive Parallel Proeessors（MPP） DeepComp 1800（Lenovo）.The testing results show that the optimization of 8×8 bijective S-boxes based on the parallel algorithm has good speedup.Meanwhile,it is validated that the master-slave parallel algorithm is suitable to solve the problem which the fitness function evaluation calculation time is much great.

关 键 词：S盒 主从模式 并行遗传算法 消息传递接口 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构]