论文部分内容阅读
针对新出现的高性能价格比的集群式计算方式 ,提出了设计高效 SPMD(single program multiple data)算法的几个原则 ,并基于这些原则 ,给出了求解多极值点优化问题的 GSAD(genetic sim ulated annealing and downhill)算法的描述。该算法有机地结合了遗传算法、模拟退火以及下山的优点 ,达到了高效、收敛、可扩展的效果。基于 MPI编程实现 ,给出了该算法在几个典型的多极值点函数以及实际问题中的应用效果 ,通过与相关工作的简单对比指出了该算法的适用范围和特色。建立 SPMD求解模型是 SPMD算法深入研究的方向
Aiming at the emerging cluster computing method with high performance-price ratio, several principles for designing single-program SPMD (single program multiple data) algorithm are proposed. Based on these principles, GSAD sim ulated annealing and downhill) algorithm description. The algorithm combines the advantages of genetic algorithm, simulated annealing and down-hill, and achieves high efficiency, convergence and scalability. Based on the MPI programming, the application of the algorithm in several typical multi-extremal point functions and practical problems is given. The application scope and features of the algorithm are pointed out by simple comparison with the related work. The establishment of SPMD solution model is the direction of further research on SPMD algorithm