Implicit conversion from float -110.125 to int loses precision