Design of Quality-Configurable Approximate Multipliers Suitable for Dynamic Environment

Approximate computing, quality configurable circuit, evolutionary algorithms, cartesian genetic programming

In the context of approximate computing, many different design approaches have been introduced in recent years exploiting the application error resilience in order to reduce power consumption. Among others, the concept of Quality Configurable Circuits (QCC) has been proposed. QCCs allow us to dynamically modulate the degree to which their functionality is approximated with respect to the desired quality of service. Existing approaches typically construct complex approximate circuits using smaller approximate and exact components. This approach is, however, associated with a considerable area and power overhead. To mitigate this issue, we propose a holistic approach enabling to formulate the QCC design problem as a search problem that can be solved by means of soft-computing techniques. In order to prove this concept, various quality configurable 8-bit multipliers were designed. Compared to the state-of-the-art solutions, the proposed quality configurable multipliers exhibit better tradeoffs in terms of electrical as well as quality parameters.