Floating-point tricks

摘要

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.