Implicit conversion from float -4.4690265486725735 to int loses precision