Implicit conversion from float -4.40894568690095 to int loses precision