Implicit conversion from float -1.9811320754716917 to int loses precision