Interval arithmetic implementations: using floating point arithmetic

This paper presents some algorithms implementing interval arithmetic using floating point arithmetic. The algorithms apply to almost any digital computer supporting normalized floating point arithmetic and provide better performance than conventional interval arithmetic program libraries. For reasons of generality and machine independence, algorithms are presented in a high-level language. They are intended to be used as an implementation guide for interval arithmetic. This paper outlines the requirements for the algorithms, presents the employed techniques and some of the algorithms, and reports on various performance experiments. A detailled description can be found in a seperate document (Clemmesen [1982]) available at the above address.