Implicit conversion from float -4.514767932489448 to int loses precision