Implicit conversion from float -32.755102040816325 to int loses precision