Implicit conversion from float -1.2844036697247674 to int loses precision