I am trying to see which features we know are necessary for mixed-state quantum computing to avoid the algorithms being efficiently simulated on classical computers.
In the case of pure-state quantum computing for instance we know that the entanglement must grow with the size of the problem we wish to simulate. We also know it is not sufficient because of the Gottesman Knill theorem.
I know that in the case of mixed-state quantum computing the situation is "more complicated". I am trying to find a relatively recent review that summarizes all that we know about the resources required in mixed-state quantum computing, especially the role of entanglement. This is the purpose of my question.