Implicit conversion from float -10.764951321279568 to int loses precision