Implicit conversion from float -2.71844660194175 to int loses precision