Implicit conversion from float -1.318681318681314 to int loses precision