I have not checked the conversation yet, but I would be surprised if it was that bad. I know that some uses of OO-features can cause a performance drop in combination with a particular compiler, but I never heard of such an impact. Performance measurements are always tricky and a bad/clumsy/careless implementation can cause bad results. For instance: iterating over an array is fast, but not OO, so an OO approach might use instead a linked-list and iterate over the elements of the list. That will definitely cause a slow-down.
Nevertheless, this is an intriguing point and I think we should have some benchmarks, like you suggest. They should be carefully constructed,