How does the presence of C++'s placement new() affect devirtualization opportunities? I'm sure Chris knows what he's talking about, so I'd like to hear more.
One interesting note is that, the last time I looked into this, the compiler _is_ allowed to assume that multiple sequential vtable calls do not replace the vtable, allowing the vtable lookup to be cached. But that is not true for some arbitrary external-to-the-compilation-unit function that.
Interestingly enough, when a use of placement new is visible to the compiler it can prevent the conservative behavior mentioned above. In https://goo.gl/Y8rwYG the vtable store that placement new conceptually generates allows Clang to devirtualize the second virtual call. GCC doesn't appear to catch this case.