Implicit conversion from float -9.478390461997009 to int loses precision