Implicit conversion from float -20.319767441860463 to int loses precision