并行算法可伸缩性的E微商分析法

The E-differetiation Method for Analyzing Scalability of Parallel Algorithms

1.引言 并行算法在并行体系结构上的可伸缩性分析(Scalability analysis)是目前巨量并行理MPP研究的中心问题之一。可伸缩性作为巨量并行机上并行算法的主要性能指标。

The E-differentiation method,which is derived from the idea of isoefficiency analysis and used for analysing scalability of parallel algorithms on parallel architectures, is proposed in this paper. E-differentiation can overcome the incompletion of other metrics for analyzing scalability which confine the manner of increasing of W w. r. t. p under some condition ;and it is easy to give the equivalent expression of these metrics in terms of E-differentiation thereof, relationships be-tween these metrics become more definitive. We conclude with an example of scalability analysis using E-differentiation and compare it with isoefficiency funcfion method.