Implicit conversion from float -5.732484076433124 to int loses precision