Implicit conversion from float -12.609308885754587 to int loses precision