SEAMBA: A Semi-Approximate Multiplier using Block-Based Approach and Rounding

Arnab Sarkar, Rajat Bhattacharjya, Ashish Goswami, Soumyajit Poddar

Abstract: Approximate computing in recent times has evolved in a humongous manner due to the extensive error-resilient capability of various data-intensive applications. In this paper, an approximate multiplier (SEAMBA) is proposed. It is called semi-approximate because it has both accurate and approximate components in its design. A block-based mechanism alongside rounding is used here so as to target specific block approximations. Experimental results show that SEAMBA is more accurate compared to other state-of-the-art approximate multipliers and it has a delay that is lesser than them by 24.7% on an average.