Implicit conversion from float -33.58642972536349 to int loses precision