Lawrence A. Crowl, "Guidelines for Presenting Parallel Performance", Technical Report 92-80-07 Revised, Department of Computer Science, Oregon State University, Corvallis, Oregon, 97331-3202, November 1993
Programmers use parallelism to improve program performance, so effective presentation of parallel performance is important. Unfortunately, common presentation practice often distorts parallel performance or requires considerable effort to interpret the results. This paper gives advice on how to measure, present, and compare parallel performance so as to reduce distortion and communicate more effectively. The primary emphasis is on visual presentation, with a secondary emphasis on data measurement.