Implicit conversion from float -20.829959514170042 to int loses precision