Implicit conversion from float -1.544117647058826 to int loses precision