Efficient Hardware Implementation of Eigen Solver

Eigen analysis is a fundamental module in many real time signal processing algorithms. This article presents the hardware architecture design and implemen-tation of a vector based high-throughput parameterizable eigen solver. The proposed design calculates an eigen values and vectors based on the QR algorithm. The QR algorithm works really well for small dense matrices and it works as additive method in projection for large matrix eigen analysis. This paper presents parallel vector based computational units for eigen solver using basic QR algorithm. The QR decomposition is implemented based on the Givens-CORDIC algorithm and the matrix multiplication of matrices$\mathrm{Q}$and$\mathrm{R}$is based on parallel vector products. The Eigen Solver is synthesized on Xilinx Virtex-7 for a real-valued matrices of size$4\times 4,8\times 8$and$16\times 16$.