clearing bits far away from computation - Printable Version +- The Orion's Arm Universe Project Forums (https://www.orionsarm.com/forum) +-- Forum: Offtopics and Extras; Other Cool Stuff (https://www.orionsarm.com/forum/forumdisplay.php?fid=2) +--- Forum: Real Life But OA Relevant (https://www.orionsarm.com/forum/forumdisplay.php?fid=7) +--- Thread: clearing bits far away from computation (/showthread.php?tid=1560) Pages: 1 2 3 4 RE: clearing bits far away from computation - Drashner1 - 07-05-2015 Anders is one of the founding members of OA actually. He hasn't been active with the project for quite some time, but his work still appears throughout the EG in many places. As you've probably already seen, there are a lot of megastructures in the OA setting. Megastructures are fun Todd RE: clearing bits far away from computation - Bear - 07-05-2015 (07-05-2015, 03:53 AM)Bob Jenkins Wrote: It wouldn't slow down computation at all. You can send one batch of bits off to be erased, and immediately start using zeros from another batch that have already been erased. No matter how you slice it though, there's no advantage. Clearing the bits locally at site A, and conducting the heat away to site B using a superconductor, will under NO circumstances be less efficient than swapping bits from site A to site B using a superconductor, clearing them producing the heat at site B, and swapping back to site A using a superconductor. So... even if it doesn't slow down computation at all, why would anybody do it? RE: clearing bits far away from computation - Bob Jenkins - 07-06-2015 (07-05-2015, 12:12 PM)Bear Wrote: No matter how you slice it though, there's no advantage. Clearing the bits locally at site A, and conducting the heat away to site B using a superconductor, will under NO circumstances be less efficient than swapping bits from site A to site B using a superconductor, clearing them producing the heat at site B, and swapping back to site A using a superconductor. So... even if it doesn't slow down computation at all, why would anybody do it? It wasn't swapping bits from A to B using a superconductor. It was packaging garbage bits into a rock at site A and throwing it to site B, B clears them, then B throwing rocks full of cleared bits back to A. The advantage (over superconductors) is that the space between A and B can be empty, instead of filled with matter. For example A being a sun-sized core and B being shell at the distance of Pluto. The advantage of packaging garbage bits instead of heat in the rocks is that heat radiates and garbage bits don't. RE: clearing bits far away from computation - Bear - 07-06-2015 If you package bits into a rock at site A and throw the rock, you're converting a bunch of energy from some source into kinetic energy - which ensures that energy being expressed as heat, defeating your purpose. In fact, I'm pretty sure you're guaranteed not to break even on this. The bits you're packaging into the rock make the rock heavier (bits are entropy, entropy is energy, E=mc^2) establishing a lowerbound on the mass required to contain them. That gives you a mass density per volume dependent on the speed with which the mass moves, putting a lowerbound on the (radiating) energy required to accelerate the mass - and it turns out that you're generating heat at the source at least equal to the heat you're "displacing" to a remote site. Same basic argument as Maxwell's Demon. RE: clearing bits far away from computation - Bob Jenkins - 07-07-2015 If you could pack bits at 1 bit per neutron or proton, a kilogram contains 6.022e26 bits. By Landauer's principle, clearing 6.022e26 bits costs at least 6.022e26 * 1.38e-23 * T * ln(2) = 28801 joules at 5 degrees Kelvin. That many joules could accelerate that 1kg to 28801 meters per second. The escape velocity from the sun is 617000 m/s, which is about 22x greater than that, so this design is a complete waste if transferring momentum to incoming to outgoing rocks consumes over 1/22 of the momentum. If it takes the weight of 10 or 100 neutrons to represent a bit, that 22x becomes 220x or 2200x. If that 617000m/s were a slower speed (like 100m/s), I can handwave how to do that. Have a big incoming rock, a stationary outgoing rock, and a very long rope hooked to the outgoing rock. Hook it onto the incoming rock as it comes by so the rope is perpendicular to the incoming rock's velocity. The whole system will go taut and rotate, bringing the incoming rock momentarily to stationary while the outgoing rock is going at the same speed the incoming rock had been coming, in the same direction. Release the rope from the outgoing rock at that instant. Tah-dah, except for the momentum stuck in the rope, which can easily be made less than 1/220 of incoming rock's original momentum. But with an escape velocity of 617000m/s I don't think a rope would work. Transferring momentum that efficiently at high speeds is something I don't know how to do. It sounds like a difficult but not necessarily impossible problem. The smaller the system, the smaller the escape velocity, and the easier this piece of the design is to solve. RE: clearing bits far away from computation - stevebowers - 07-08-2015 Can orbital rings help this problem? Paul Birch wanted to use them for everything, from megastructure surface support to spacecraft launch and planetary cooling. Here's a red hot planet being actively cooled by a series of eccentric loops. Perhaps you could couple your waste data pellets to a series of rings. RE: clearing bits far away from computation - Bob Jenkins - 10-07-2015 Orbital rings! I hadn't encountered that before. I'll have to think about those. My immediate reaction is that anything tethered to the ring at a speed different from the ring will find some variant of friction pulling it along with the ring, which can be overcome by using power for stationkeeping. Also questions about the strength needed to hold it together, but that's probably well understood and just requires homework on my part. Meanwhile I found a different solution for throwing rocks. If you have moon-sized lumps of heavy stuff like iron or lead in orbit around the core, close passes to those moons can cause large changes in velocity. If you have an equal mass of rocks speeding up and slowing down, they'll cancel out and the heavy moon can maintain a normal orbit. If one pass isn't enough, have several such moons at appropriate distances. That way the first toss only has to reach the gravitational influence of the first moon, it doesn't have to get all the way to the outer shell on its own. I was also worried about an outer shell (of whatever shape) billowing like an overly large soap bubble in the wind. Because any point of a shell is essentially a flat thin sheet, and can't push or pull against perpendicular forces. You could instead compensate in a sufficiently large ring around the billow, where there's been enough curvature that you can push along the shell rather than perpendicular to the shell. Any number of such compensations can be summed, so each point can push or pull appropriately to correct any billowing across the whole structure. You could punch the shell anywhere successfully, but the shell could always fix it relatively quickly. RE: clearing bits far away from computation - Bear - 10-08-2015 It's true though, that displacing the mass using mechanical energy generated far from the mass is definitely viable. So you can have a distant source of energy helping to remove energy from the local frame and the objection I raised a couple posts ago can be sidestepped. RE: clearing bits far away from computation - Tachyon - 10-11-2015 (06-05-2015, 06:26 PM)Bob Jenkins Wrote: Irreversible computing can be done as reversible computing plus bit clearing. Swapping bits is reversible. You don't have to clear bits at the same place that computation consumes zeros. You could clear bits someplace safe to generate heat, toss a bunch of bits to the computation, swap zeros for random bits, then toss the random bits back to be cleared again. Does that help with making computation compact and low temperature? Not really. I'm not sure what you mean by "swapping bits", but writing a bit requires $\log_e(2)k_BT$ joules per bit, and is an irreversible process. If you don't write the result of a (classical) computation, you cannot know it. If we are talking qubits, then reading the result state involves collapse of the wave function, which is also inherently non-reversible. Attempts to evade this by weak measurement run into the problem that until the result is observed, the solution doesn't exist (or rather exists as a multiplicity of possible outcomes which don't allow the quantum algorithm to proceed further). Even if you wave your hands and invoke Many-Worlds, wave function collapse is still irreversible in our particular branch of reality. When you write "clear bits someplace safe" you are really discussing information transfer rate, which in temperatures ranging from 10E3 to 10E9K is (in bits per second): $\dot{I}\le\frac{4(A\sigma_B)^{1/4}P^{3/4}}{3k_{B}\log_e(2)}$ So you haven't fooled nature or otherwise gotten around fundamental limits of computation. https://cloud.sagemath.com/#projects/970500fb-5f43-4c43-b127-129dd10dadae/files/limits-of-computation.sagews As discussed maybe a decade ago on the old Orion's Arm Yahoo lists, reversible computing allows you to reduce the power expenditure of computing an N-step algorithm by $N^{b}$ for some 0 (x, x XOR y) -> (x XOR x XOR y, x XOR y)==(y, x XOR y) -> (y, x XOR y XOR y)==(y, x). That's swapping bits. If you want zeros in the second bit, it's enough to be able to irreversibly clear the first bit. An improvement in another part of this design: the energy and accuracy needed to toss rocks large distances. It can be aided by gravitational assists. And gravitational assists don't need accurate tosses up front, you can split the rock into two or three pieces in flight so that each piece gets a gravitational assist by different objects. If there is an equal mass of rocks going out as coming in, the accelerating assists in one direction are cancelled by the decelerating assists in the other, so the large objects doing the assist pretty much go along their orbit unhindered. I'm still stuck on the mass distribution / stability of the central core. I'm getting closer to being able to simulate it, I have a high precision floating point library and interpolation working now.