On: GNU GCC does not round floating-point divisions to the nearest value

[saagarjha]

(Note that in your example no accuracy is lost)