Implicit conversion from float -12.933673469387756 to int loses precision