Implicit conversion from float -3.209876543209873 to int loses precision