Implicit conversion from float -12.964426877470359 to int loses precision