Implicit conversion from float -381.2041884816754 to int loses precision