Implicit conversion from float -1.3705583756345163 to int loses precision