ResearchBlog‎ > ‎

[Papers] Parallel Slack

posted Aug 16, 2011, 2:41 PM by Teng-Yok Lee   [ updated Aug 16, 2011, 8:11 PM ]
Parallel: a metric to evaluate the benefit of parallelization for the improvement of performance. Two realted articles are listed below:

J. K. Hollingsworth and B. P. Miller. Slack: A New Performance Metric for Parallel Programs. 1992.

Critical Path Profiling is a technique that provides guidance to help programmers try to improve the running time of their program. However, Critical Path Profiling provides only an upper bound estimate of the improvement possible in a parallel program execution. In this paper, we present a new metric, called Slack , to complement Critical Path and provide additional information to parallel programmers about the potential impact of making improvements along the critical path.

J. K. Hollingsworth and B. P. Miller. 1992. Parallel program performance metrics: a comprison and validation. In Proceedings of the 1992 ACM/IEEE conference on Supercomputing (Supercomputing '92), Robert Werner (Ed.). IEEE Computer Society Press, Los Alamitos, CA, USA, 4-13.

The authors present a novel technique, called true zeroing, that permits direct, quantitative, and fair comparison of parallel program performance metrics. This technique was applied to three programs that include both numeric and symbolic applications. Three existing metrics, Gprof, Critical Path, and Quartz/NPT, and several new variations were compared. The result of this comparison was that while Critical Path provided the best overall guidance, it was not universally better than the other metrics. Because there is no single universal metric, future parallel performance systems need to support multiple metrics. The authors present a set of recommendations to tool builders based on the experience gained during this case study