Abstract
We have implemented the GEN BLOCK (generalized block) data distribution in PGHPF, our High Performance Fortran implementation. Compared to a BLOCK or CYCLIC distribution, the more flexible GEN BLOCK distribution allows users to balance the load between processors. Simple benchmark programs demonstrate the benefits of the new distribution format for unbalanced work loads, getting speedup of up to 2X over simple distributions. We also show performance results for a whole application.
Chapter PDF
References
High Performance Fortran Language Specification, Version 2.0 (1997)
Hu, Y.C., Johnnson, S.L., Teng, S-H., High Performance Fortran for Highly Irregular Problems, Proc. Sixth Symp. Principles and Practices of Parallel Programming, ACM Press (1997) 13–24
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1999 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Leair, M., Miles, D., Schuster, V., Wolfe, M. (1999). Flexible Data Distribution in PGHPF. In: Amestoy, P., et al. Euro-Par’99 Parallel Processing. Euro-Par 1999. Lecture Notes in Computer Science, vol 1685. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-48311-X_55
Download citation
DOI: https://doi.org/10.1007/3-540-48311-X_55
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-66443-7
Online ISBN: 978-3-540-48311-3
eBook Packages: Springer Book Archive