Implicit conversion from float -21.989100817438683 to int loses precision