Implicit conversion from float -36.30573248407643 to int loses precision