Implicit conversion from float -5.062836624775585 to int loses precision