Implicit conversion from float -5.912408759124091 to int loses precision