Integrating Task and Data Parallelism by means of Coordination Patterns
Manuel Diaz(University of Malaga
), Bartolome Rubio, Enrique Soler, Jose M. Troya
To appear at (HIPS'01), San Francisco, California, USA, April 23, 2001
Abstract
This paper shows, by means of some examples, the suitability and expressiveness of
a pattern-based approach to integrate task and data parallelism. Coordination
skeletons or patterns express task parallelism among a collection of data parallel
HPF tasks. Patterns specify the interaction among domains involved in the
application along with the processor and data layouts. On the one hand, the use
abstract of domains, i.e. regions together with some interaction information, improves
pattern reusability. On the other hand, the knowledge at the coordination level
of data distribution belonging to the different HPF tasks is the key for an efficient
implementation of the communication among them. Besides that, our system implementation
requires no change to the runtime system support of the HPF compiler used. We also
present some experimental results that show the efficiency of the model.
