Topic: Green computing (Read 206 times)

Hydrogen-I propose that the first real world application of energy efficient reversible computing should be something that is both simple and compatible with reversibility. Linear feedback shift register are pretty much the simplest things that do a non-trivial (but still characteristic 2 linear) calculation. But we probably need non-linearity to do a more useful calculation, so we will need to use a non-linear feedback shift register NLFSR.

Before you worry about the processor for reversible computation, you should look into the algorithms and maybe write a few reversible programs.

Here are a couple of problems for you to work on:

1. Given an easily computable function f, how might one construct a reversible combinatorial circuit to compute x->(x,f(x))? Do not worry about time/space efficiency. Just perform the calculation.

2. Given an easily computable invertible function f and its inverse g, how might one construct a reversible combinatorial circuit to compute x->f(x) (you will need to use 1 and g for this one)?

Have you heard of Bennett's pebble game? Perhaps you should play that game. You may also play a generalized version of Bennett's pebble game on a directed graph.

You can try writing reversible programs in the language Janus here: https://topps.diku.dk/pirc/?id=janus

Let me know when you have followed these basic instructions, and then we can continue.

How would you propose to implement "reversible computing"?

Compression? Optimization? Better branch prediction? Improved multi threading? Bigger symmetric multi processor? Bigger cache?

Not only is there not a buzzword for what you're proposing. It lacks an instruction set.

digaran-I cannot respond to the content of your post unless someone quotes you because you are a shithead who has absolutely nothing good or interesting to say. You are worthless.

Can you please live long and stay around? I mean if you continue like this who will be left to quote the last person for you to see their post? As far as I can see you are ignoring anyone talking to you. Lol

I just quoted franky for you, now you need someone to quote me for you to see our posts, lets hope the next person is not ignored by you.😂

franky1-I am ignoring you. If you want me to respond to the content of your post rather than just its existence, then you need to get someone to quote you. You are a chlurmck.

-Joseph Van Name Ph.D.

seems the OP has not learned basic lessons of social communication and computing.
computers already do reverse computing. hard drives are partitioned with factory settings. software settings have restore points. programs have "undo" (ctrl-z) options. files are not deleted, they are preserved and overwritten when deemed useless(FAT), thus saving on any deletion energy

its not the future, its the present

There have been proposals for using superconductors for energy efficient reversible computing. Since my expertise is in pure mathematics, I have a limited understand of the proposals. But high temperature superconductors do not seem to be very useful for much of anything, so if we are going to use low temperature superconducting reversible computers, we should get ready to spend a ton of energy (which is a lot of energy since e=mc^2) on cooling the superconductors to a temperature where the superconductor performs well (which is likely below the maximum superconducting temperature).

The few people who know about energy efficient reversible computing will agree that it is difficult, so it is probably best to solve a problem that I would call "reversible computing advantage". This is the problem of creating an energy efficient reversible computer that can solve a specific task more energy efficiently than a conventional irreversible computer. If we can make that specific task profitable, then we have officially entered the age of energy efficient reversible computing (aka green computing).

Ofcourse, it seems the benefit of energy efficient computer outweighs the amount of memory it requires to run a reversable computation assuming you don't have to deal with another energy issue like electrical resistance which researchers on superconductivity seek to solve. Atleast, reversable computation could in theory reduce energy loss  rather than eliminate it completely

Ucy-Reversible computation is the only way to get energy efficiency. Energy efficiency (below Landauer's limit) and more powerful computation is the end goal, and reversible computation is the only way to attain this goal.

But with clever algorithms, one can manage the garbage information so that one does not accumulate garbage information while computing. For example, if you can compute both a function f and its inverse g, one can transform an input x into f(x) completely reversibly without producing any garbage information. If you can compute a function h which is not necessarily invertible, then you can reversibly transform an input x to the pair (x,h(x)) without producing any garbage information (the mapping x->(x,h(x)) is injective). Through variations of Bennett's pebble game, one can transform any irreversible algorithm into a reversible algorithm with a small algorithmic overhead. Irreversible time T and space S becomes reversible time O(T^(1+epsilon)) and space O(S*(1+log(T/S))) for any epsilon>0. The computational complexity overheads are manageable. And you can get even better results by using partial reversible computation. By trading quantities such as space,time, parallelism, and reversibility, one can optimize the performance of the computation using pebble games, but to obtain the best tradeoff, one will need physically reversible computational hardware.

Reversible computation requires more memory. And memory is getting cheaper and cheaper and more and more abundant. Right now, the main bottleneck to better computation is the restrictions imposed by Landauer's principle. And reversible computation is the only way to get past this bottleneck. We can always add more memory. But we cannot make atoms smaller, and we cannot change Boltzmann's constant.

Since reversible computation requires a little bit more space and time (but not too much more space and time), one should expect for reversible computation to greatly outperform irreversible computation in the long-term, but in the near future, it will probably be quite difficult to make energy efficient physically reversible computers that have an advantage in a specific problem over irreversible computers.

-Joseph Van Name Ph.D. (in pure math)

Seems energy efficiency is not the only goal of this project but to be able to run a computation in reverse. I assume the only computation worth preserving and reversing is non garbage computation. Does  it also preserve garbage or computation that is of no value?  Won't that be a problem on scarce computer memory?

Reversible computing is the art of computing while deleting very little information. Said differently, reversible computing is the art of computing where one can recover the initial state of the computation from the final state of the computation by running the computation in reverse. Landauer's principle states that deleting 1 bit of information costs k*T*ln(2) energy where k is Boltzmann's constant (k=1.38*10^(-23)Joules/Kelvin), but in practice deleting a bit of information must cost somewhere around >>100*k*T just to overcome thermal noise and reliably delete the information. On the other hand, there is no theoretical non-zero minimum amount of energy required for reversible computation where one does not delete information. This means that reversible computation will become much more energy efficient than irreversible computation. Since reversible computing will eventually become much more energy efficient than the conventional irreversible computation that we have today, reversible computers will eventually greatly outperform irreversible computers.

So why have people not heard of reversible computing? Why do they immediately and angrily dismiss reversible computing? It is not because reversible computing is infeasible. There are plenty of infeasible ideas that are greatly overhyped with obnoxious timescales. It is simply because reversible computing does not have a buzzword. People love AI because it has a buzzword. People love nuclear fusion because 'nuclear fusion' is a buzzword. People do not care whether it is infeasible. The same thing goes for green energy, quantum computing, and anything associated with 'science' (baby formula and covid vaccines are soooooooo scientific according to CNN). One reason why reversible computing is not popular is because there is no buzzword associated with reversible computing and because the media which tells you all exactly what to think (because you cannot think for yourselves) has not pumped up reversible computing.

This is why I propose the term 'green computing' as a buzzword to refer to reversible computing. Unfortunately, the media will have a really hard time gobbling up the term 'green computing' when referring to reversible computing. This is because the media is like a toilet. The media likes the taste of bullshit and only gobbles up bullshit. It is very difficult to convince the media of anything legitimate because they are like kids in the candy store with out their mommies and daddies, and they reject nutrition in favor of artificially manufactured lumps of sugar.

Reversible computing is the future. Green computing is the future.