Effective Processor Architecture for Matrix Decomposition

Fast Implementation for the Singular Value and Eigenvalue Decomposition Based on FPGA

“…We show the computation time can be reduced significantly because the Diagonal processors (DPs) and Non-diagonal processors (NDPs) have almost 100% hardware utilization. In addition, we perform the EVD with a simplified CORDIC-like algorithm [16] , which broadcasts the sign set rather than rotation angles, and leads to an improvement in area and time.…”