Check the largest recovery error

A computer has a finite number of bits for each stored numerical value. Many decimal fractions therefore cannot be represented exactly in binary floating-point storage. Repeated arithmetic must round intermediate results.