Implicit conversion from float -10.048828125 to int loses precision