Abstract
A three-dimensional mesh implementation of Csanky’s method is shown to be asymptotically faster than systolic algorithms for n × n full matrix inversion. This requires the fastest known sequential matrix multiplication algorithms with large hidden constants. It nevertheless suggests the possibility of o(n) matrix inversion on a scalable architecture with realistic communication costs. Three-dimensional fan-in requirements imply a lower bound of Ω(n ½) or Ω(n 2/3) depending on the input’s spatial distribution. The example of inversion illustrates general methods for optimal mesh implementation of matrix algorithms.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
D. P. Bertsekas and J. N. Tsitsiklis, Parallel and Distributed Computation. Numerical Methods, Prentice-Hall International, 1989.
D. Coppersmith and S. Winograd, Matrix multiplication via arithmetic progressions, in Proceedings 19th ACM STOC, New-York, 1987, pp. 1–6.
L. Csanky, Fast parallel matrix inversion algorithms, SIAM Journal on Computing, 5 (1976), pp. 618–623.
W. M. Gentleman, Some complexity results for matrix computations on parallel processors, Journal of the ACM, 25 (1978), pp. 112–115.
A. M. Gibbons and Y. N. Srikant, A class of problems efficiently solvable on mesh-connected computers including dynamic expression evaluation, Information Processing Letters, (1989), pp. 305–311.
M. Kunde and T. Tensi, Multi-packet-routing on mesh connected arrays, in ACM Symposium on Parallel Algorithms and Architectures, Santa Fe, New-Mexico, June 1989, pp. 336–343.
R. E. Ladner and M. J. Fischer, Parallel prefix computation, Journal of the ACM, 27 (1980), pp. 831–838.
S. Lakshmivarahan and S. K. Dhall, Analysis and Design of Parallel Algorithms. Arithmetic and Matrix Problems, McGraw-Hill, 1990.
D. Nassimi and S. Sahni, Data broadcasting in SIMD computers, IEEE transactions on computers, c-30 (1981), pp. 101–107.
M. C. Pease, Inversion of matrices by partitioning, Journal of the ACM. 16 (1969), pp. 302–314.
F. P. Preparata and D. V. Sarwate, An improved parallel processor bound in fast matrix inversion, Information Processing Letters, 7 (1978), pp. 148–150.
P. M. B. Vitànyi, Locality, communication, and interconnect length in multicomputers, SIAM Journal on Computing, 17 (1988), pp. 659–672.
J. G. G. V. D. Vorst, Solving the least squares problem using a parallel linear algebra library, Future Generation Computer Systems, 4 (1988-1989), pp. 293–297.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1991 Springer Science+Business Media New York
About this chapter
Cite this chapter
Hains, G. (1991). Matrix Inversion in 3 Dimensions. In: Mullin, L.M.R., Jenkins, M., Hains, G., Bernecky, R., Gao, G. (eds) Arrays, Functional Languages, and Parallel Systems. Springer, Boston, MA. https://doi.org/10.1007/978-1-4615-4002-1_16
Download citation
DOI: https://doi.org/10.1007/978-1-4615-4002-1_16
Publisher Name: Springer, Boston, MA
Print ISBN: 978-1-4613-6789-5
Online ISBN: 978-1-4615-4002-1
eBook Packages: Springer Book Archive