Implicit conversion from float -2.1126760563380174 to int loses precision