Implicit conversion from float -6.615720524017476 to int loses precision