Implicit conversion from float -5.701943844492433 to int loses precision