Implicit conversion from float -2.503863987635242 to int loses precision