Fun fact, the state update function can be modelled as the multiplication of the state vector by a constant matrix (where the vector/matrix elements are all in GF(2)). By raising that matrix to a power n, you can use the resultant matrix to compute n state updates at once, allowing you to "skip ahead" arbitrarily into the output stream.

Hypothetically, you could use this to vectorize/parallelize computation of xorshift128+ output sequences in large batches, but I'm not sure anyone's ever bothered to do this because the naive sequential method is so cheap.