Energy Efficient Compact Approximate Multiplier for Error-Resilient Applications

The primary goal of approximate computing is enhancing system performance, such as energy efficiency, speed, and form factor. Despite the growing use of approximate multipliers, the design of efficient approximate compressors – a fundamental multiplier block – remains a significant challenge. In this paper, 8-transistor and 14-transistor 4:2 compressors are proposed. Both compressors exploit CMOS technology and a constant and conditional approximation of selected inputs, exhibiting fewer negative errors. As a result, a resource-expensive error recovery module is eliminated, yielding superior performance as compared with prior art. The 14-transistor architecture yields a lower error rate compared to the 8-transistor architecture, trading off lower area for higher accuracy. The compressor-tailored circuit architecture is also proposed and evaluated using image multiplication. The proposed multiplier exhibits 50% area savings and 93% lower power-delay-product compared to the exact multiplier, as well as higher accuracy, and 38% PDP enhancement compared with the state-of-the-art.