Floating-point arithmetic with 84-bit numbers

Abstract
A classic and straightforward technique is presented which is not limited to the size or type of number representation used or multiple precision arithmetic.