Implicit conversion from float -15.291005291005291 to int loses precision