Implicit conversion from float -2.965779467680605 to int loses precision