Heuristic algorithms of communication optimization in parallel PGAS-programs
Abstract
This paper represents the heuristic algorithms of communication optimization in parallel PGAS-programs providing time minimization of its execution. This is achieved by accounting of hierarchical structure of computer systems while executing reduction and preloading of remote arrays to computer system's nodes. Developed algorithms are implemented for PGAS languages: Cray Chapel and IBM X10 and simulated on cluster computer systems.
Keywords
PGAS,
Cray Chapel,
IBM X10,
PGAS,
compiler optimization,
reduction,
scalar replace,
Cray Chapel,
IBM X10
About the Authors
I. .. Kulagin
СибГУТИ
Russian Federation
Russian Federation
A. .. Paznikov
СибГУТИ
Russian Federation
Russian Federation
M. .. Kurnosov
СибГУТИ
Russian Federation
Russian Federation
References
1. Хорошевский В.Г. Распределённые вычислительные системы с программируемой струк-турой // Вестник СибГУТИ. 2010. № 2 (10). С. 3-41.
2. Rabenseifner R. Optimization of Collective Reduction Operations // Computational Science - ICCS 2004 - Lecture Notes in Computer Science. 2004. Vol. 3036. P. 1-9.
3. Li S., Hoefler T., Snir M. NUMA-Aware Shared-Memory Collective Communication for MPI // HPDC 2013. 2013. P. 85-96.
4. Курносов М.Г. Алгоритмы трансляционно-циклических информационных обменов в иерархических распределённых вычислительных системах // Вестник компьютерных и информационных технологий. 2011. № 5. С. 27 - 34.
5. Kennedy K., Allen John R. Optimizing compilers for modern architectures: a dependence-based approach. Morgan Kaufmann Publishers Inc. San Francisco, CA, USA 2002. 834 p.
6. Barik R., Zhao J., Grove D., Peshansky I., Budimlic Z., Sarkar V. Communication Optimizations for Distirbuted - Memory X10 Programs // IEEE International Parallel and Distributed Processing Symposium. 2011. P. 1 - 13.
7. Chen W., Iancu C., Yelick K. Communication Optimizations for Fine-grained UPC Applications // 14th International Conference on Parallel Architectures and Compilation Techniques (PACT), 2005. Tech Report LBNL-58382.
8. Mallón D.A., Taboada G.L., Teijeiro C., Domínguez J.G., Gómez A., Wibecan B. Scalable PGAS collective operations in NUMA clusters // Cluster Computing. 2014. 23 p.
9. Nishtalab R., Zhenga Y., Hargrovea P.H., Yelick K.A. Tuning collective communication for Partitioned Global Address Space programming models // Parallel Computing. 2011. Vol. 37. P. 576 - 591.
10. Callahan D., Chamberlain B.L., Zima H.P. The Cascade High Productivity Language // HIPS 2004. 2004. P. 52 - 60.
11. Nanjegowda R., Hernandez O., Chapman B., Jin H. Scalability Evaluation of Barrier Algorithms for OpenMP // IWOMP '09 Proceedings of the 5th International Workshop on OpenMP: Evolving OpenMP in an Age of Extreme Parallelism. 2009. Vol. 5568. P. 42 - 52.
12. Callahan D., Carr S., Kennedy K. Improving register allocation for subscripted variables // PLDI ’90, New York, NY, USA, 1990. ACM. P. 53 - 65.
13. Charles P., Donawa C., Ebcioglu K., Grothoff C., Kielstra A., Praun C., Saraswat V., Sarkar V. X10: An Object-oriented approach to non-uniform Clustered Computing // OOPSLA 2005. P. 519 - 538.
Review
For citations:
Kulagin I..., Paznikov A..., Kurnosov M... Heuristic algorithms of communication optimization in parallel PGAS-programs. The Herald of the Siberian State University of Telecommunications and Information Science. 2014;(3):52-66. (In Russ.)
Views:
181
Email this article 
