This avoids another float computation, avoiding a potential source of
rounding issues
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>