A comparative analysis between logarithmic number system and floating-point ALU

Abstract

The ease and accuracy of executing the multiplication and division operations by using either fixed-point addition or subtraction is what makes the logarithmic number system an attractive option. However, its main drawback is the complexity of performing addition and subtraction operations. In this paper, an analysis is carried out to compare the current state-of-the-art logarithmic number system and floating-point. The results predominantly show that the logarithmic number system offers better speed, accuracy and power-efficiency than floating-point. A new proposal for logarithmic addition and subtraction using a co-transformation method is presented, with the aim of improving its performance further.