Implicit conversion from float -36.04477611940297 to int loses precision