Implicit conversion from float -5.637583892617442 to int loses precision