A systolic architecture for the singular value decomposition

75. R. P. Brent and F. T. Luk, A systolic architecture for the singular value decomposition, Report TR-CS-82-09, DCS, ANU; Report TR 82-522, DCS, Cornell University; Report CMA-R02-82, ANU, August 1982, 12 pp.


We propose a systolic architecture for computing a singular value decomposition of an m × n matrix, where > n. Our algorithm is stable and requires only O(mnS) time on a linear array of O(n) processors. Here S is a slowly growing function of n; for practical values of n we may regard S as a constant. Extensions to algorithms for two-dimensional arrays are also discussed.


A revision of this report is incorporated in [84].

Go to next publication

Return to Richard Brent's index page