Implicit conversion from float -10.341796875 to int loses precision