作者: Tom Deakin , Jason Sewall , S. John Pennycook , Douglas Jacobsen , and Simon McIntosh-Smith
DOI: 10.1109/P3HPC51967.2020.00007
关键词: Statistical analysis 、 Software portability 、 Software metric 、 Computer science 、 Performance efficiency 、 Critical assessment 、 Visualization 、 Measure (data warehouse) 、 Data science
摘要: Recent work has introduced a number of tools and techniques for reasoning about the interplay between application performance portability, or "performance portability". These have proven useful setting goals guiding high-level discussions, but our understanding portability problem remains incomplete. Different views same efficiency data offer different insights into an application's (or lack thereof): standard statistical measures such as mean deviation require careful interpretation, even metrics designed specifically to measure may obscure differences applications.This paper offers critical assessment existing approaches summarizing across platforms, proposes visualization means extract information underlying distribution. We explore alternative visualizations, outlining new methodology that enables developers reason their applications how it might be improved. This study unpicks what portable" provides question.