Abstract

Loading and storing a matrix associated with equations defining a physical system into the memory of a multiprocessor is accomplished by scanning along one orthogonal axis of the matrix and loading non-zero elements into memory associated with a plurality of the processors to form a transpose vector, with diagonal elements of the matrix being stored in the memory of a selected one of the processors. Each stored matrix element is provided with an index which identifies the origin of the element along the other axis of the matrix.

Other References

• "The Finite Element Method for Engineers" by Huebner et al., John Wiley and Sons, 1982, Chapters 1-3
• "Introduction to Numerical Computations" by Vandergraft, Academic Press, Second Edition, Chapter 6
• "Elementary Numerical Analysis" by Conte et al., McGraw-Hill, 1972, Section 3.2
• "Computational Models for Future Signal Processing Supercomputers" by Kung, Dept. of Computer Science, Carnegie Mellon University, Pittsburgh, PA 8/4/86
• "Wafer-Scale Integration and Two-Level Pipelined Implementations of Systolic Arrays" by Kung et al., Journal of Parallel and Distributed Computing, No. 1, 1984, pp. 32-63
• "A Systolic Array Computer" by Arnould et al., IEEE Int'l Conference on Acoustics, Speech, and Signal Processing, Mar. 26-29, 1985
• "Inner Product Computers" by Swartzlander, Jr. et al., IEEE Trans. on Computers, vol. C-27, No. 1, Jan. 1978, pp. 21-31
• "Systolic Arrays [for VLSI]" by Kung et al., Sparce Matrix Proceedings 1978, Society for Industrial and Applied Math., pp. 256-282
• "An Efficient Parallel Algorithm for the Solution of Large Sparse Linear Matrix Equations" Arnold et al., IEEE Trans. on Computers, vol. C-32, No. 3, Mar. 1983, pp. 265-272