Implicit conversion from float -3.8005390835579647 to int loses precision