I notice that the Wikipedia article refers here to "thermodynamically reversible" which is perhaps not the same thing as computationally reversible. So I looked up "thermodynamically reversible" and found
At the level we're talking about, the distinction between thermodynamics and computational theory gets a little hazy. Seriously -- we're talking about the deepest magics of math and physics, where in order to get peak efficiency from our CPUs we have to exploit the Bekenstein limits of the event horizons of black holes. So, "perhaps not the same thing" is absolutely true. Perhaps a lot more similar than we think. At this level of theory, things get pretty wacky.
I studied quantum computation and quantum information theory some in grad school. My takeaway from it was that by the standards of the field I am basically a dog who's learned to shake hands, sitting at a table listening to Deutsch and Witten and Susskind argue and thinking that I'm really smart just because I can recognize one word every few minutes. Every now and again they look over my way, realize I'm paying attention, say "good boy!" and scratch my ears and I bark and think I'm making a real contribution to the discussion.
Yes, I am *that far* out of my depth here -- Scott Aronson I ain't. Please be careful about thinking I'm any kind of an authority here.
which gives the interesting summary: thermodynamically reversible processes are theoretical and don't occur in the real world.
Yep. Second Law again: entropy must always increase, meaning nothing is truly thermodynamically reversible. But if adiabatic computing *did* exist, man, it would be cool -- as I said, if someone's able to demonstrate it I'll be deeply fascinated. (And then I'll try to see if I can leverage it to travel back in time. Because hey, once the Second Law no longer limits you, the world's your oyster.)
That article seems confused as to whether a reversible process must be infinitely slow or infinitely fast, but Wikipedia says the former: http://en.wikipedia.org/wiki/Reversible_process_%28thermodynamics%29
The closer you approach energy-free computing, the slower the process goes -- this is a consequence of several different things, including the Margolus-Levitin theorem, which says that a bit can't be flipped faster than h/4E seconds. The less energy you put in, the slower the flip goes.
_______________________________________________ Gnupg-users mailing list Gnupg-users@gnupg.org http://lists.gnupg.org/mailman/listinfo/gnupg-users
