Inexact computing with approximate adder application

Abstract

This paper presents an analysis of an inexact N-bit ripple carry adder architecture. Results show that a 30 percent power reduction is achieved for several approximate adders while maintaining a root-mean square error of 16 percent.