Implicit conversion from float -6.448087431693992 to int loses precision