Implicit conversion from float -592.7710843373494 to int loses precision