Implicit conversion from float -20.696517412935336 to int loses precision