Floating-point tricks
J.F. Blinn · IEEE Computer Graphics and Applications · 1997
The author discusses IEEE floating point representation that stores numbers in what amounts to scientific notation. He considers the sign bit, the logarithm function, function approximations, errors and refinements.