A VLSI inner-product processor for real-time DSP applications

摘要

A VLSI design of a 12/spl times/12-bit parallel inner-product processor (IPP) for two's complement multiplication is presented. The designed architecture and its driving algorithm makes an efficient use of all the array cells in order to reduce the computation time. Therefore, the time-area complexity of the designed IPP is proportional to N/spl times/m/sup 2spl times/T/sub FA/ for large N and is m times smaller than for a sequential computation on a single multiplier. An inner product of two N-dimensional vectors with m-bit words requires area proportional to m(m/2), and the delay time of (N+2m+log/sub 2/N)/spl times/T/sub FA/, where T/sub FA/ denotes the delay of a full adder. The design has been fabricated in 2-/spl mu/m CMOS double-metal technology. An example application of the designed IPP for solution of a triangular system is presented.