I think this is a really good idea -- you guys have an interesting proof, and implementation results to back it up. It would also be a good place to talk about the IL and its effectiveness at expressing/exposing optimizations. -- Suresh