Implicit conversion from float -4.09937888198759 to int loses precision