eps

Floating-point relative accuracy (machine epsilon).

Syntax

y = eps
y = eps(X)

Description

Returns the distance from 1.0 to the next larger double-precision number — about 2.2204e-16. With argument X, returns the distance from X to the next representable number of the same type. Useful for tolerance-based comparisons.

Examples

eps                      % 2.2204e-16
abs(0.1 + 0.2 - 0.3) < eps     % 1  (tolerance check)
eps(1e10)                % ≈ 1.86e-6  (gap at 1e10)

See also

• realmin — Smallest positive normalized floating-point number.
• realmax — Largest positive finite floating-point number.
• pi — Mathematical constant π.