Machine epsilon

The rounding unit of the floating-point system: the maximum relative error of representing a number.

Definition

It is the gap between 11 and the next representable floating-point number. In double precision (IEEE 754) it equals 2522.22×10162^{-52}\approx2.22\times10^{-16}, which translates into about 15–16 significant decimal digits.

How it is used

It sets the floor for any reasonable tolerance: asking for a relative error below machine epsilon makes no numerical sense.