Implicit conversion from float -21.755725190839712 to int loses precision