Implicit conversion from float -32.13930348258708 to int loses precision