I think the async benchmark could be faster still when pinned to a single core if a single threaded runtime was used, and possibly if a single-thread channel im

I think the async benchmark could be faster still when pinned to a single core if a single threaded runtime was used, and possibly if a single-thread channel implementations were used, but then it's becoming a bit academic. Really, what async gives you is a programming style that's very similar to using blocking sockets, but allows one to achieve select() like performance when doing I/O. That, and it allows one to not have to have a special thread for timers or, even worse, a thread per timer, as that's hidden away by the async runtime implementation and just works.