What will be the computation accuracy if both x and y can be stored(exactly) in floating-point in the system.
$x$ - $\\frac{x}{y}*y$
