Abstract
Efficient parallelization and optimization of programs on distributed memory parallel systems demands a considerable amount of knowledge about the underlying target machine and parallelizing compiler. Clearly, an automatic performance estimator would significantly alleviate the parallelization effort by relieving the programmer from the burden of studying machine and compiler details.
This paper describes how to automatically estimate the critical network contention behavior as induced by a parallel program on a parallel architecture. The approach assumes a hypercube network topology and a fixed and static routing policy. Careful modeling of the virtual to physical processor mapping, communication implementation, network topology and the underlying routing policy allows to prevent expensive simulation of the network contention behavior. The method is therefore low in computational complexity and memory requirements. A proof is presented which demonstrates the absence of network contention for an important class of communication statements.
The network contention cost function is fully implemented as part of the P 3 T, which is a static parameter based performance prediction tool under the Vienna Fortran Compilation System (VFCS).
Preview
Unable to display preview. Download preview PDF.
References
D. Atapattu and D. Gannon. Building Analytical Models into an Interactive Performance Prediction Tool. In Proc. Supercomputing 89, pages 521–530, Reno, Nevada, 1989. ACM Press.
V. Balasundaram, G. Fox, K. Kennedy, and U. Kremer. A Static Performance Estimator to Guide Data Partitioning Decisions. In 3rd ACM Sigplan Symposium on Principles and Practice of Parallel Programming (PPoPP), Williamsburg, VA, April 21–24 1991.
L. Bhuyan, Q. Yang, and D. Argawal. Performance of multiprocessor interconnection networks. IEEE Computer, pages 25–37, Februar 1988.
S. Bokhari. A network flow model for load balancing in circuit-switched multicomputers. NASA Contractor Report 182049, Institute for Computer Applications in Science and Engineering, NASA Langley Research Center, Hampton, VA 23665, May 1990.
S. Bokhari. Communication overhead on the intel iPSC-860 Hypercube. ICASE Interim Report 10, NASA Contractor Report 182055, Institute for Computer Applications in Science and Engineering, NASA Langley Research Center, Hampton, VA 23665, May 1990.
S. Bokhari. Complete Exchange on the iPSC-860. ICASE Report No. 91-4, NASA Contractor Report 187498, Institute for Computer Applications in Science and Engineering, NASA Langley Research Center, Hampton, VA 23665, January 1991.
B. Chapman, S. Benkner, R. Blasko, P. Brezany, M. Egg, T. Fahringer, H.M. Gerndt, J. Hulman, B. Knaus, P. Kutschera, H. Moritsch, A. Schwald, V. Sipkova, and H.P. Zima. VIENNA FORTRAN Compilation System — Version 1.0 — User's Guide, January 1993.
T. Fahringer. Automatic Performance Prediction for Parallel Programs on Massively Parallel Computers. PhD thesis, University of Vienna, Department of Software Technology and Parallel Systems, October 1993.
T. Fahringer. The Weight Finder, An Advanced Profiler for Fortran Programs. In Automatic Parallelization, New Approaches to Code Generation, Data Distribution, and Performance Prediction. Vieweg Advanced Studies in Computer Science, ISBN 3-528-05401-8, Verlag Vieweg, Wiesbaden, Germany, March 1993.
T. Fahringer and H. Zima. A Static Parameter based Performance Prediction Tool for Parallel Programs. In Invited Paper, In Proc. of the 7th ACM International Conference on Supercomputing 1993, Tokyo, Japan, July 1993.
G. Fox, M. Johnson, G. Lyzenga, S. Otto, J. Salmon, and D. Walker. Solving Problems on Concurrent Processors, volume 1. and 2. Prentice Hall, Englewood Cliffs, NY, 1988.
H.M. Gerndt. Parallelization for Distributed-Memory Multiprocessing Systems. PhD thesis, University of Bonn, December 1989.
C. Kruskal and M. Snir. The performance of multistage interconnection networks for multiprocessors. IEEE Transactions on Computers, C-32:1091–1098, Dec. 1983.
F.H. McMahon. The Livermore Fortran Kernels Test of the Numerical Performance Range, pages 143–186. Elsevier Science B.V., North Holland, Amsterdam, 1988.
G. Pfister and V.A. Norton. Hot Spot Contention and Combining in Multistage Interconnection Networks. In Proc. of the 1985 International Conference on Parallel Processing, pages 790–797, 1985.
W.H. Press, B.P. Flannery, S.A. Teukolsky, and W.T. Vetterling. Numerical Recipes in C; The Art of Scientific Computing. Cambridge University Press, 1988.
D. Russel. The Principles of Computer Networking. Cambridge University Press, Cambridge, Great Britain, 1991.
V. Sarkar. Partitioning and Scheduling Parallel Programs for Multiprocessor. The MIT Press, Cambridge, Massachusetts, 1989.
A. Sethi and N. Deo. Interference in multiprocessor systems with localized memory access probabilities. IEEE Transactions on Computers, C-28(2), Feb. 1979.
K.Y. Wang. A Performance Prediction Model for Parallel Compilers. Technical report, Computer Science Dept., Purdue University, November 1990. Technical Report CSD-TR-1041, CAPO Report CER-90-43.
H. Zima, P. Brezany, B. Chapman, P. Mehrotra, and A. Schwald. Vienna Fortran — a language specification. Technical report, ICASE, Hampton,VA, 1992. ICASE Internal Report 21.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1994 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Fahringer, T. (1994). Automatically estimating network contention of parallel programs. In: Haring, G., Kotsis, G. (eds) Computer Performance Evaluation Modelling Techniques and Tools. TOOLS 1994. Lecture Notes in Computer Science, vol 794. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-58021-2_22
Download citation
DOI: https://doi.org/10.1007/3-540-58021-2_22
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-58021-8
Online ISBN: 978-3-540-48416-5
eBook Packages: Springer Book Archive