Implicit conversion from float -20.543840177580464 to int loses precision