Implicit conversion from float -6.556886227544908 to int loses precision