Approximate Compressors for Multiplication

Abstract

At nanometric scales, approximate computing is an attractive prototype used for digital processing. Despite providing less accurate results, approximate computing is preferred over exact computing as it provides a fast & significant output along with low power consumption. Designing of an efficient multiplier has always been a challenge for VLSI designers as multipliers have a large area, long latency consumes considerable power. For this inconvenience compressor with low latency, low power consumption and reduced stages of the product are designed. This paper proposes two methods to design high order compressors (8:4 & 9:4) (i) Using adders (half & full) (ii) Using multiplexers in Cadence VIRTUOSO tool using 45nm technology. Extensive simulation results show that the proposed designs achieve significant accuracy improvement along with power, area, and delay reductions compared to previous compressor designs.