Not having read this particular discussion about the connection between EMH and P=NP, I would have thought that EMH was much stronger. Just as you can come up with market problems that are NP-complete, it seems to me that you should be able to come up with non-Turing-computable market problems, so that full efficient markets would be required to solve uncomputable problems. Perhaps there's a natural way to reduce the logical strength needed by weakening the EMH, but I should think about this.