Implicit conversion from float -12.657807308970092 to int loses precision