A CORDIC Processor Array for the SVD of a Complex Matrix

Abstract

Matrix factorizations are important in many real-time signal processing applications. In order to improve the performance of these algorithms, special-purpose VLSI processor arrays are being developed. Recently, the Coordinate Rotation Digital Computer (CORDIC) algorithms have been applied to the QR Decomposition (QRD) and the Singular Value Decomposition (SVD). In this paper, the CORDIC arithmetic algorithms are extended to deal with complex data. CORDIC VLSI architecture for the QRD of a complex matrix, the Eigenvalue Decomposition of a Hermitian matrix, and the SVD of a complex matrix are presented and the area and time complexity of the processor elements are analyzed. VLSI implementation of a CORDIC SVD processor element is under development at Rice University.