Implicit conversion from float -32.63231197771587 to int loses precision