Implicit conversion from float -101.57427937915742 to int loses precision