Did you cite the right paper? There is nothing about doing computation at that energy level or a rate of 100mhz, the whole paper insteads seem to be about how one can more effectively simulate mechanical computational devices on current hardware.
Here is the other (more main paper)
https://arxiv.org/pdf/1801.03534.pdf. No. This is not about current devices because we currently do not have those energy efficient reversible computers. We are instead looking to directions to innovate.
But practically, how would reversible computing work? To avoid burning energy on erasing data, you'd have to keep the result of every operation. How would caching work, a technology that is vital for keeping our computers fast and entirely based on erasing data when it needs space for new data?
Trick 0: When using reversible computation, one should use partial reversibility instead of complete reversibility. Landauer's limit is not a lot of energy, so to get the most out of reversible computation, one needs to find the sweet spot between irreversibility and reversibility.
Trick 1: Suppose that one would like to compute a function f on the input a. Then using a reversible computer, by storing all of the bits generated in the computation, after the computation, one would obtain f(a) along with G(a) which is the garbage information that we generate. In other words, our computation performs the transformation
a->(f(a),G(a)). Now, we would like to get rid of the garbage information G(a). To do this, we would first copy our desired output f(a) to obtain a->(f(a),G(a),f(a)). We would then run our computation a->(f(a),G(a)) in reverse to transform (f(a),G(a)) back into a. When we put everything together, we have computed the mapping
a->(a,f(a)) reversibly. The function a->(a,f(a)) is injective and it is a restriction of the bijection (a,b)->(a,f(a) XOR b) which can also be performed on a reversible computer quite easily.
Trick 2: Suppose that f,g are inverse functions. Then as long as we are able to reversibly transform a to (a,f(a)) and b to (b,g(b)), we may also reversibly transform a to f(a) without producing any garbage information. In particular, we perform the following transformations a->(a,f(a))=(g(f(a)),f(a))->f(a).
Trick 3: One can iterate Trick 1 by using Bennett's pebble game (perhaps generalized to partial reversible computation and digraphs) in order to compute with a manageable computational complexity theoretic overhead.
Trick 4: One can also design functions including encryption and hashing functions for (partial) reversibility.
I have to remind all of you that NIST has butchered cryptographic functions such as SHA-256 and AES by standardizing low quality irreversible cryptographic functions rather than the reversible ones (or partially reversible in the case of SHA-256). I have to give NIST a grade of F- for this egregious oversight.
What would your computer do once you've streamed a full movie online, and have generated terabytes of raw screen data? Does pixels on the screen not "count" in your energy budgets, since they are outside the CPU? They are still erased dozens of times per second, millions of them.
-It is called partial reversibility. We do not have to immediately use reversibility everywhere in order for reversibility to be applicable. Besides, the lowest frequency of visible light has an energy of about 75 kT per photon, so if your device is emitting photons, it is losing more than Landauer's limit of energy per photon.
And if you have any evidence, post it here in this thread for all to see.
-You are an extraordinarily arrogant chlurmcklet. Go away.
yeah he doesn't know how to answer a question like that. he just believes... but i'll give it a shot. one of the issues it seems to me is the one you bring up that it would increase the complexity of the computer's hardware architecture to have to "keep the result of every operation". probably not cost effective. you don't get something for nothing but let's hear what Mr. Ph.D has to say. he's the authority.
-You are contributing absolutely nothing to the conversation because you are a very low quality specimen.
he won't know how to answer that question i'm pretty sure unless he just links you to some summary of some research paper that they did to avoid losing tenure...
since I'm the OP I wish i could put him onto a moderation in this thread to kind of tamper his enthusiasm for "reversible computing". but i guess we don't have that feature.
-You have that nasty attitude because you hate science and you hate research. You are contributing absolutely nothing to the conversation. And unlike a good steak, the meat on your bones has been completely wasted. How sad! And partial reversibility is a thing too. Please learn how to read what I have been telling you. But you can't do that because you hate science. I have much more respect for flat-Earthers than I do for specimens like you. Since you are so pathetic, I am pressing the ignore button. I will not respond to your bullshit anymore unless someone quotes you. I am also ignoring digiran because that entity is also a chlurmck who is not worth talking to.
-Joseph Van Name Ph.D.
below a couple of reference (one in Italian language).