Implicit conversion from float -16.53973509933776 to int loses precision