Implicit conversion from float -5.412844036697237 to int loses precision