Parallelisation of desktop environments

Abstract

The objective of this research was to investigate the parallelisation of desktop environments in light of multi-core processors becoming mainstream. In contrast to the parallelisation of typical high performance computer applications, parallel computing for desktop applications involves further challenges. Although improving performance is the primary aspect of parallel computing, a vital focus of this research was on the software engineering approach. In particular, this included developing an object-oriented solution to desktop parallelisation. To address these issues, two concepts were developed in this research. The first concept, the Parallel Iterator, is an object-oriented solution for data parallelism. The Parallel Iterator is the parallel extension to the standard sequential iterator, to support parallel traversal of elements in an arbitrary collection. The second concept, Parallel Task, is an object-oriented solution for task parallelism. An important design aspect of Parallel Task was its semantic integration with the structure of typical desktop applications. Both of these concepts have been successfully implemented. The solutions provide ease of use by following familiar programming approaches and encapsulating parallelisation details away from programmers. The results for both the Parallel Iterator and Parallel Task concepts show superior performance compared to standard parallelisation approaches. This research has made a vital contribution to parallel computing on mainstream desktop systems. By identifying challenges specific to the parallelisation of desktop applications and their current structure, the concepts developed are in line with object-oriented programming and the software engineering approach. The concepts and tools developed not only ease the programming effort, but also enhance the user's desktop experience by promoting responsive and interactive applications.