A Low-Power and Small-Area Multiplier for Accuracy-Scalable Approximate Computing

Many applications are error resilient in that they can tolerate some level of inaccuracy in their computations. Approximate computing benefits such applications by trading power for accuracy. The priority between power and accuracy depends on each application. Some prefer low power, whereas others require high accuracy. This paper focuses on the former and proposes a very-low-power and small-area multiplier for accuracy-scalable approximate computing. The proposed approximate multiplier consists of an approximate tree compressor (ATC) and a carry-maskable adder (CMA). The ATC compresses partial products power-efficiently with a simple circuit structure, whereas the CMA provides accuracy scalability using a simple carry-masking technique. From the experimental designs, it is unveiled that the proposed approximate multiplier reduces power consumption and circuit area by 68.8% and 61.9%, respectively, in the most accurate configuration compared with the Wallace tree multiplier.