Implicit conversion from float -104.5754716981132 to int loses precision