Abstract
The general block distribution of a matrix is a rectilinear partition of the matrix into orthogonal blocks such that the maximum sum of the elements within a single block is minimized. This corresponds to partitioning the matrix onto parallel processors so as to minimize processor load while maintaining regular communication patterns. Applications of the problem include various parallel sparse matrix computations, compilers for high-performance languages, particle in cell computations, video and image compression, and simulations associated with a communication network.
We analyze the performance guarantee of a natural and practical heuristic based on iterative refinement, which has previously been shown to give good empirical results. When p 2 is the number of blocks, we show that the tight performance ratio is θ(√p). When the matrix has rows of large cost, the details of the objective function of the algorithm are shown to be important, since a naive implementation can lead to a Ω(p) performance ratio. Extensions to more general cost functions, higher-dimensional arrays, and randomized initial configurations are also considered.
In comparison, Khanna et al. have shown the problem to be approximable within O(1) factor [6], while Grigni and Manne have shown it to be NP-hard to approximate within any constant factor less than two [4].
Preview
Unable to display preview. Download preview PDF.
References
B. Aspvall, M. M. Halldórsson, and F. Manne, Approximating the Generalized Block Distribution of a Matrix, Institutt for Informatikk, TR-141, University of Bergen, Dec. 1997.
B. Chapman, P. Mehrotra, and H. Zima, Extending HPF for advanced data parallel applications, IEEE Trans. Par. Dist. Syst., (Fall 1994), pp. 59–70.
G. N. Frederickson, Optimal algorithms for partitioning trees and locating p-centers in in trees, in Proceedings of second ACM-SIAM Symposium on Discrete Algorithms, 1991, pp. 168–177.
M. Grigni and F. Manne, On the complexity of the generalized block distribution, in Proceedings of Irregular'96, the third international workshop on parallel algorithms for irregularly structured problems, Lecture Notes in Computer Science 1117, Springer, 1996, pp. 319–326.
High Performance Fortran Language Specification 2.0, January 1997. Available from http://www.crpc.rice.edu/HPFF/home.html.
S. Khanna, S. Muthukrishnan, and S. Skiena, Efficient array partitioning. Proceedings of the 24th International Colloquium on Para Automata, Languages, and Programming (ICALP), Lecture Notes in Computer Science 1256, Springer, 1997, pp. 616–626.
F. Manne, Load Balancing in Parallel Sparse Matrix Computations, PhD thesis, University of Bergen, Norway, 1993.
F. Manne and T. Sørevik, Partitioning an array onto a mesh of processors, in Proceedings of '96, Workshop on Applied Parallel Computing in Industrial Problems and Optimization, Lecture Notes in Computer Science 1184, Springer, 1996, pp. 467–477.
A. Mingozzi, S. Ricciardelli, and M. Spadoni, Partitioning a matrix to minimize the maximum cost, Disc. Appl. Math., 62 (1995), pp. 221–248.
D. M. Nicol, Rectilinear partitioning of irregular data parallel computations, J. Par. Dist. Comp., (1994), pp. 119–134.
B. Olstad and F. Manne, Efficient partitioning of sequences, IEEE Trans. Cornput., 44 (1995), pp. 1322–1326.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1998 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Aspvall, B., Halldórsson, M.M., Manne, F. (1998). Approximations for the general block distribution of a matrix. In: Arnborg, S., Ivansson, L. (eds) Algorithm Theory — SWAT'98. SWAT 1998. Lecture Notes in Computer Science, vol 1432. Springer, Berlin, Heidelberg. https://doi.org/10.1007/BFb0054354
Download citation
DOI: https://doi.org/10.1007/BFb0054354
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-64682-2
Online ISBN: 978-3-540-69106-8
eBook Packages: Springer Book Archive