Implicit conversion from float -0.6293706293706407 to int loses precision