Implicit conversion from float -2.696629213483149 to int loses precision