Implicit conversion from float -1.395348837209312 to int loses precision