Implicit conversion from float -2.696629213483135 to int loses precision