Implicit conversion from float -5.33898305084746 to int loses precision