Published Paper Details:

The great majority of contemporary computer systems use the IEEE Standard for Floating-Point Arithmetic (IEEE 754), which has been the industry standard for floating-point arithmetic for many years. Recently, posit (Type III unum), a new number representation format, has been suggested as an alternative to the widely used IEEE 754 arithmetic. John L. Gustafson says this new format may deliver superior accuracy utilizing equal or less bits and simpler hardware than current standard. This undergraduate thesis analyses and contrasts the innovative posit number format's qualities and properties with the accepted practice for floating-point numbers (floats). We concentrate on assessing if posits would be a good "drop-in replacement" for floats based on claims made in the literature. First we propose a low-level design for posit arithmetic multiplier using the Xilinx tool to generate synthesizable HDL code which helps in the case of only unsigned numbers of multiplication. Where as in the practical, we need to focus on both signed and un-signed numbers. So here we proposed a new technique called RoBA (Rounding Based Approximate) multiplier which helps in reducing the area, delay and power by 10%, 40% and 54% respectively. To conclude this work, we propose a low-level design for posit arithmetic (signed and un-signed) RoBA multiplier using the Xilinx tool to generate synthesizable HDL code. Designed using XilinxISE14.7 software.