Implicit conversion from float -21.0944527736132 to int loses precision