Enhancing data parallel aplications with task parallelism

Abstract

Most parallel applications contain data parallelism and almost all discussion of its solutions has limited to the simplest and least expressive form: flat data parallelism. Several generalization of the flat data parallel model have been proposed because a large number of those applications need a combination of task and data parallelism to represent their natural computation structure and to achieve good performance in their results. Their aim is to allow the capability of combining the easiness of programming of the data parallel model with the efficiency of the task parallel model. In this work, we examine how to enhance two basic data parallel computation applications with task parallelism. Applications presented: N-body Simulation and Echo Elimination Process have been chosen from an unlimited scope of applications where the benefit of the integration of task and data parallelism can be shown