Implicit conversion from float -6.131386861313871 to int loses precision