Implicit conversion from float -3.080357142857139 to int loses precision