Mixed task and data parallel executions in general linear methods

Article type: Research Article

Authors: Rauber, Thomas | Rünger, Gudula

Affiliations: University Bayreuth, Germany | Chemnitz University of Technology, Germany. E-mail: ruenger@informatik.tu-chemnitz.de

Note: [] Corresponding author. University Bayreuth, Angewandte Informatik II, 95440 Bayreuth, Germany. Tel.: +49 921 555100; Fax: +49 921 555102; E-mail: rauber@uni-bayreuth.de

Abstract: On many parallel target platforms it can be advantageous to implement parallel applications as a collection of multiprocessor tasks that are concurrently executed and are internally implemented with fine-grain SPMD parallelism. A class of applications which can benefit from this programming style are methods for solving systems of ordinary differential equations. Many recent solvers have been designed with an additional potential of method parallelism, but the actual effectiveness of mixed task and data parallelism depends on the specific communication and computation requirements imposed by the equation to be solved. In this paper we study mixed task and data parallel implementations for general linear methods realized using a library for multiprocessor task programming. Experiments on a number of different platforms show good efficiency results.

Keywords: General linear methods, ordinary differential equations, task parallelism, data parallelism, orthogonal structures of communication

Journal: Scientific Programming, vol. 15, no. 3, pp. 137-155, 2007