Implicit conversion from float -7.45054945054946 to int loses precision