摘要
具有巨量数据的科学计算程序在并行计算机上的有效速度往往是很低的,因此需要尽可能地借助算法来改善这种状况。基于高性能科学计算中对某些基本线性计算模型的处理要耗费大部分CPU时间以及cache具有较高速度这一事实,文章着重对这些计算模型的cache利用率CUR(CacheUsingRatio)进行了分析与实验比较,同时也对用户如何获取程序的浮点运算速度的方法进行了讨论。
The effective speed of the high-performance scientific computing program with huge data on parallel comput-er is very slow,so it is necessary to improve this state of affairs with the help of algorithm as much as possible.In view of the fact that the basic linear calculation models expend most CPU time in high-performance scientific comput-ing and cache has higher operating speed in hierarchical memory,the paper lays stress on the analysis of CUR of these calculation models and on comparing their testing results,at the same time ,it also discusses applied determination method of obtaining float operating speed of parallel program in running.
出处
《计算机工程与应用》
CSCD
北大核心
2002年第17期135-137,共3页
Computer Engineering and Applications
基金
国家社会科学基金项目(编号:97BTQ004)
