4. Parallel Programming Models

Abstract

Previous chapter Next chapter Software, Environments, and Tools Parallel MATLAB for Multicore and Multinode Computers4. Parallel Programming Modelspp.55 - 76Chapter DOI:https://doi.org/10.1137/1.9780898718126.ch4PDFBibTexSections ToolsAdd to favoritesExport CitationTrack CitationsEmail SectionsAboutExcerpt Summary A nonembarrassingly parallel example (Blurimage) is used to illustrate the key design, code, debug, and test trade-offs associated with parallel programming. The Blurimage example reveals the breadth of issues associated with applying these steps. In addition, this chapter will also provide a short introduction to the primary parallel programming models: distributed arrays, message passing, and manager/worker. This chapter focuses on providing more breadth regarding the key design, code, debug, and test trade-offs associated with parallel programming. The goal is to reveal the breadth of issues associated with applying these principles. In addition, this chapter will also provide a short introduction to the primary parallel programming models: distributed array semantics, message passing, and manager/worker. Some of the principles covered in this section include the following: Design: when to go parallel, classes of parallel programs, parallel programming models, machine model Coding: different parallel programming styles, impact of code on performance and scalability Debug: details of the parallel debug process, what errors are found at each step Testing: performance metrics and predicting performance (relative to machine parameters) The primary example program used in this chapter is more sophisticated than those shown in Chapter 2 and is designed to expose some of the boundaries of the different parallel programming models. 4.1 Design: Knowing when to “go parallel” The single most important decision in parallel programing is knowing when a parallel computer is required. Previous chapter Next chapter RelatedDetails Published:2009ISBN:978-0-89871-673-3eISBN:978-0-89871-812-6 https://doi.org/10.1137/1.9780898718126Book Series Name:Software, Environments, and ToolsBook Code:SE21Book Pages:xxii + 242Key words:Parallel computing, MATLAB, Multicore, simulation, parallel programming