Implicit conversion from float -36.59176029962548 to int loses precision