Implicit conversion from float -32.118155619596536 to int loses precision