ABSTRACT
Unums represent the entire real number line with a limited number of bits, and make it possible to define a computing environment that does not round, overflow, or underflow. Just as floats incorporate scaling information into the number to make it easier to do math on computers, unums take this idea a few steps further by incorporating the exact or inexact status of a number, the number of bits of precision, and the dynamic range. In practice, floats are grossly over-sized as insurance against things that can ruin the result. Because unums automatically scale the bits needed up and down, they usually use fewer bits than floats, yet produce more accurate answers that come with provable bounds.
