Implicit conversion from float -21.994609164420496 to int loses precision