Implicit conversion from float -2.47933884297521 to int loses precision