Modules for Teaching Parallel Performance Concepts

• Published in: Topics in Parallel and Distributed Computing pp. 59 - 77
• Main Author: Qasem, Apan
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing 2018
• Subjects: OpenMP; Parallel Performance Metrics; Parallel Program Execution; Super-linear Speedup; Task Granularity
• ISBN: 9783319931081; 3319931083
• DOI: 10.1007/978-3-319-93109-8_3

This chapter introduces three teaching modules centered on parallel performance concepts. Performance related topics embody many fundamental ideas in parallel computing. In the ACM/IEEE curricular guidelines (ACM2013), an entire knowledge unit has been devoted to parallel performance. In addition, performance topics pervade every knowledge area within PDC and can be found across other knowledge areas including Algorithms, Architecture and Systems Fundamentals. The three modules presented in this chapter cover a range of parallel performance topics. Since power savings have become an important consideration from hand-held devices to supercomputers, energy efficiency is also emphasized in each module. The modules focus more on architectural and algorithmic issues rather than the programming aspects. The modules are constructed to illustrate parallel performance issues primarily through code examples and experimental studies. This approach makes the modules accessible to students who do not yet have a strong background in parallel programming. Thus, the target audience for this chapter are instructors who are teaching CS1, with or without parallel programming, and also instructors who are teaching upper-level electives where their students may already have taken a semester of parallel programming.