摘要
现有选择性计算机性能评价方法主要使用基准程序评价方法,基准程序中各子程序的输出往往因为单位不同而无法进行进一步数据处理,同时基准程序评价方法广泛采用的权重和评分方法缺乏理论依据。针对该问题,提出基于佩奇排名(PageRank)的计算机性能评价方法,采用比较数据序列间相似性的方法产生邻接矩阵从而为各项评估功能计算PageRank得分。实验结果表明,该方法能客观反映目标计算机系统的性能。
Benchmark programs are widely used for selection evaluation of computer systems. Since the output data have different units, it is difficult to do further data processing. Another problem about benchmark programs is that the weighted sum scores do not have a theory basis. This paper proposes a PageRank-based computer performance evaluation method, the similarity between data serials are evaluated to produce the adjacency matrix. PageRank is then used on this matrix to produce PageRank score for each target function. Experimental result demonstrates the effectiveness and more objective of the method.
出处
《计算机工程》
CAS
CSCD
北大核心
2010年第17期286-287,290,共3页
Computer Engineering
关键词
佩奇排名
基准程序
性能评价
PageRank
benchmark programs
performance evaluation
