Análise comparativa da precisão em ponto flutuante dos padrões IEEE 754 e Posit.

Abstract

Increasing technological development has demanded more and more precision and efficiency in arithmetic calculations. Recently a new floating-point representation of numbers was proposed to replace the conventional IEEE-754 standard. The present work is based on the comparison of floating-point accuracy of IEEE-754 standard and posit when using the same bit depth. In order to make this comparative analysis, the Fast Fourier Transform was calculated for discrete-time signals. The accuracy of the results generated with float and posit formats were determined relative to the 64 bits data type.