Implicit conversion from float -20.878378378378372 to int loses precision