Implicit conversion from float -21.64519326065411 to int loses precision