Using Template Matching to Infer Parallel Design Patterns

摘要: The triumphant spread of multicore processors over the past decade increases pressure on software developers to exploit growing amount parallelism available in hardware. However, writing parallel programs is generally challenging. For sequential programs, formulation design patterns marked a turning point development, boosting programmer productivity and leading more reusable maintainable code. While literature now also reporting rising number patterns, programmers confronted with task parallelizing an existing program still struggle question which pattern apply where their In this article, we show how template matching, technique traditionally used discovery can be support parallelization decisions. After looking for matches previously extracted dynamic dependence graph, classify code blocks input according structure find. Based information, easily implement detected create version his or her program. We tested our approach six successfully pipeline do-all patterns.