ABSTRACT
The universal number, or unum, encompasses all standard floating point formats, as well as fixed point and exact integer arithmetic. Unums get more accurate answers than floating point arithmetic, yet use fewer bits in many cases, which saves memory, bandwidth, energy, and power. Unlike floating point numbers (“floats”), unums make no rounding errors, and cannot overflow or underflow. That may seem impossible for a finite number of bits, but that is what unums do. Unums are to floats what floats are to integers: A superset.
