Integer Representations in IEEE 754, Posit, and Takum Arithmetics

Abstract: The posit and takum machine number formats have been proposed as alternatives to the IEEE 754 floating-point standard. As floating-point numbers are frequently employed to represent integral values, with certain applications explicitly relying on this capability, it is pertinent to evaluate how effectively these new formats fulfil this function compared to the standard they seek to replace. While empirical results are known for posits, this aspect has yet to be formally investigated. This paper provides rigorous derivations and proofs of the integral representation capabilities of posits and takums, examining both the exact number of bits required to represent a given integer and the largest consecutive integer that can be represented with a specified number of bits. The findings indicate that, while posits are generally less effective than IEEE 754 floating-point numbers in this regard, takums demonstrate overall superior representational strength compared to both IEEE 754 and posits.