Implicit conversion from float -1.9104991394147959 to int loses precision