use float constants instead of double as computations