Implicit conversion from float -12.763496143958875 to int loses precision