The function takes a numerical input x and scales it by a factor of around 2, but with a different sign depending on the input. It also applies some rounding or truncation, which causes the output to be noisy.

