Implicit conversion from float -8.626817447495952 to int loses precision