I see, thanks. I wanted to ask just to know what kind of demand we had from our users for certain algorithms we might implement in the future. For that specific one, I should mention that it's not obvious to me how it could be parallelized. A lot of the operations are dependent on results of previous operations, so it's very far from the so-called "embarassingly parallel" ideal case. But at a minimum, one could always try more multi-core parallelism of the tensor contractions themselves, and that is something we are working on soon which will help many different algorithms. Best regards, Miles