Implicit conversion from float -19.45406125166444 to int loses precision