Implicit conversion from float -0.5386740331491637 to int loses precision