Energy efficient Gaussian filtering for multimedia applications using novel approximate adders

摘要

The approximate computing has gained wide popularity to achieve low power and high-performance designs for portable devices employing multimedia applications. This paper presents a novel approximate full adder (AFA) cell by simplifying the sum and carry Boolean expressions. The proposed AFA requires a small number of logic gates with a very small probability of error. Further, novel energy-efficient approximate ripple carry adder (RCA) and carry look-ahead (CLA) adder architectures are presented that utilize proposed AFA cells. The proposed approximate RCA (ARCA) and approximate CLA (ACLA) adders are evaluated based on their implementation complexity and probability of error. The design metrics of proposed and existing adders are extracted by implementing all designs and synthesizing using Synopsys Design Compiler with 65 nm PDK. The synthesis results show that the proposed 16-bit ARCA and ACLA require 13.82 and 7.38 reduced power respectively over the existing approximate adder. Finally, the Gaussian smoothing filters (GSFs) embedded with proposed ARCA and ACLA consume 18.16 and 24.4 reduced energy compared to the GSF with existing approximate adder.