Implicit conversion from float -36.85121107266434 to int loses precision