Twenty years of research on RNS for DSP: Lessons learned and future perspectives

In this paper, we discuss a number of issues emerged from our twenty-year long experience in applying the Residue Number System (RNS) to DSP systems. In early days, RNS was mainly used to reach the maximum performance in speed. Today, RNS is also used to obtain power-efficient (tradeoffs speed-power) and reliable systems (redundant RNS). Advances in microelectronics and CAD tools play an important role in favoring one technology over another, and a winning choice of the past may become at disadvantage today. In this paper, we address a number of factors influencing the choice of RNS as the winning solution. From technology platforms (ASIC and FPGA), to issue related to modern design tools, from cost of memory, to cost of wiring, from power dissipation to thermal issues. Moreover, we mention how RNS can be helpful in implementing reliable architectures (fault detection and correction) in future VLSI systems.