Hypercube Architecture For Singular Value Decomposition And Other Fast Transforms

H. S. Hou

doi:10.1117/12.942804

Translator Disclaimer

19 February 1988 Hypercube Architecture For Singular Value Decomposition And Other Fast Transforms

H. S. Hou

Author Affiliations +

Proceedings Volume 0848, Intelligent Robots and Computer Vision VI; (1988) https://doi.org/10.1117/12.942804
Event: Advances in Intelligent Robotics Systems, 1987, Cambridge, CA, United States

Abstract

The hypercube architecture is a form of concurrent processing that uses many tightly coupled processors connected in an N-dimensional cube. It can be a multiple-instruction or a single-instruction and multiple-data machine. Using a three-dimensional cube as an example for visual convenience, this paper describes the algorithms for performing the singular value decomposition (SVD), the fast Fourier transform (FFT), the fast Hartley transform (FHT), and the cosine transform on this 3-D cube architecture. Because these algorithms when implemented on a hypercube require only the nearest neighborhood communications, not only is the communication overhead greatly reduced, but the architecture becomes modular. An additional advantage is the programming flexibility. This paper demonstrates that the same hypercube configuration can be used to process such algorithms as SVD, FFT, FHT, and cosine transforms.

Citation Download Citation

H. S. Hou "Hypercube Architecture For Singular Value Decomposition And Other Fast Transforms", Proc. SPIE 0848, Intelligent Robots and Computer Vision VI, (19 February 1988); https://doi.org/10.1117/12.942804

ACCESS THE FULL ARTICLE