A systemic performance estimation and moduli set selection method for residue number systems

Abstract

The residue number system (RNS) has been receiving considerable attention for many decades owing to its inherent carry-free and parallel properties in addition, subtraction, and multiplication operations. In RNSbased systems, the method by which the moduli set is selected and evaluation of its performance are important issues. A systemic performance evaluation method for RNS based on the properties of the moduli set is proposed in this paper. By abstracting the inherent properties of moduli sets, such as the complexity of arithmetic units, utilization ratio of dynamic range, parallelism, and balance between residue channels, information can be provided on moduli set selection and performance estimation before circuit implementation. Furthermore, we also propose a new multi-channel moduli set that utilizes a new radix component in this paper. Performance analysis and comparison results show that the proposed multi-channel moduli set has good systemic performance.