ISSN (e): 2250 – 3005 || Volume, 12 || Issue, 3|| May. - June. – 2022 || International Journal of Computational Engineering Research (IJCER) www.ijceronline.com Open Access Journal Page 9 A Study of POSIT Arithmetic Implementations Meesala Sowmya, Kala S, Nalesh S 1 Undergraduate Student, 2 Assistant Professor Department of Electronics And Communication Engineering, Indian Institute of Information Technology Kottayam, Kottayam –626635, Kerala, India. 2 Assistant Professor, Department of Electronics, Cochin University of Science And Technology, Kochi-686620, Kerala, India. Corresponding Author: Kala S --------------------------------------------------------------------------------------------------------------------------------------- Date of Submission: 13-06-2022 Date of acceptance: 27-06-2022 --------------------------------------------------------------------------------------------------------------------------------------- I. INTRODUCTION Floating-point numbers, like scientific notations, are represented by an exponent (normally in base two) and a significand, except that the significand must fit on a limited number of bits. Floating point representation is conceptually similar to scientific notation, and it consists of a signed number, also known as the significand, mantissa, coefficient, or ambiguously fraction. This number is encoded as a digit string of a specific length. It also consist of a signed integer exponent (base two) that changes the magnitude of a number. To find out the value of a floating-point number F, the significand or mantissa M is multiplied by the base β raised to the power of the exponent E, as indicated by, F = M × β E IEEE-754 floating point numbers have the format shown in Fig.1. The number contain following fields: one sign bit S, 2 bit biased exponent, the biased exponent has trailing bits indicating fractions and the leading bit of the fraction is implicitly encoded. Fig. 1: Floating point Number Format ABSTRACT Since mankind started counting, there have been multiple numbering systems, each with its own advantages and disadvantages. We use Arabic numerals, which are base-10 system, whereas computers use binary numbers, which are base-2. Representing real numbers accurately and efficiently on devices that can handle only discrete and finite information is challenging. The IEEE standard for floating point numbers is the most common implementation that modern computing systems have adopted. But due to the deficiencies in the standard, different implementations of IEEE-754 are not guaranteed to give same answers. IEEE-754 also suffer from limitations like redundant representation for numbers, signed zeros, overflow/underflow issues, etc. Floating point numbers are known to have large energy and area footprints when implemented in hardware. To address the shortcomings, a new data type called posit was proposed in 2017. Posit numbers are the result of decades of work in creating a viable replacement for floating point numbers, and they have better accuracy, speed, and simpler design. Posit arithmetic has been introduced as a replacement to IEEE 754- floating point arithmetic number system. In this paper we discuss various posit arithmetic implementations in the literature. KEYWORDS: Floating point number system, posit arithmetic, hardware implementations, accuracy.