Uploady II wrote:Well hi, it’s me. These kinds of posts will have to replace comments, at least for now. If you’re wondering how Gemini 2.0 is going, well, it has been prematurely cancelled at the puffer stage, along with a sample destruction for it. That is because of a rather genius idea by, KoffKoff, Me et al. An idea that could truly make single-channel-90 a thing of the past: (Sorry for my sensationalism)
Stable Storage Spaceship Challenge: Winning Entry
We were thinking about how to circumvent the "Stable storage? No more single-channel for you!" problem. Our first idea was unbelievably inefficient and suboptimal. But then, we came up with this:
(0, 0, 0, 0, 0...)hd Madness
TLDR: We thought “Why don’t just combine a whole bunch of these slow glider streams together to get one big one?” And, the rest is history.
So, one idea was to have a whole bunch of these weightship tape readers, turning blocks into gliders, feeding back into a whole bunch of inserters, and making one big stream. But, thst had a few blaring issues with it:
- Weightships are orthogonal, and gliders are diagonal! It was still fixable by using weightships instead of gliders for the back-signal.
There wasn’t any tape duplication!. This was still fixable, but only by taking an expensive “second pass”, one of a few ways.
Finally, the worst problem of all: It would be a rake, not a spaceship! The only ways to fix this, at least that we could think of, are all pretty overcomplicated.
We probably missed a few details, but we considered it unworkable, and moved on.
Idea “Stealing” (so we didn’t have to waste out time innovating)
A few ideas later, we came across the idea to copy the methods used by some of the fastest static-tape replicators in engineered CA. That is to have a sliding-head tape reader continuously output signals into the construction and tape duplication parts of the replicator, and therefore avoid quadratic slowdown. This fixes all of the above problems. Yay?
The Nitty Gritty Details (Finally!) (Caveat: Example only. Please make a smaller version. Mine is a real nightmare.)
This thing starts as a body, with a few tapes sticking out of it. Nothing happens, not even oscillation (this is stable), until...
- A glider starts a p256 gun, which causes all the tape readers to repeatedly DINCREAD.
The intermittent glider streams from the tape readers, of which there are a shocking 256, get split into two paths:
The first path goes to a whole bunch of tape writers, one for each reader, which do the job of writing a duplicate of the tape. Keep in mind all of the reads are destructive, so the tape isn’t duplicated, it’s just moved.
The second path, well, every one of those paths are combined into one path, in such a way that the first tape means 0 mod 256 gliders, the second 1 mod 256, and so on to 255 mod 256. That one stream can do single channel, 0hd, (0, 0)hd, and so on, as long as the inserters can cope.
You might have noticed that there is no plain p256 gun, just a “core” somewhat similar to the part on the 0E0P metacell with a p256 gun. Well after construction is completed, the whole thing self-destructs, leaving no trace except for a perfect copy of itself, shifted by some distance. The self-destruct mechanism also emits the trigger glider just before it, itself, self-destructs.
And finally, we have a static-tape self-constructing spaceship, nicknamed the Super-slow Shootable Ship for how easy it is to synthesize, and how slow it goes.
Something a little lot more optimized
We could really just get away with two readers, two writers, one core, and one decoder, to do 0hd or 10hd construction, and hugely save on size, cost, slowness, and general lag. Despite losing single-channel features, this will probably be way cheaper for reasonable step sizes, and probably far less laggy for unreasonable ones. It will also be less tedious to assemble.
Details and Further Optimization
Actually, it may be possible to use just 1 tape, which contains 8-bit numbers representing the glider spacings of the pairs in a 0hd or 10hd glider stream, with special values representing, say, a single glider. That still gives you a big toolkit, but with just a single tape. Also, a low-population variant using the spacing between blocks is possible as well, but depending on the interpretation, might be quite inefficient.
Freedom (Conclusion)
Best part is, we no longer have to single-channel everything! We could use more efficient construction methods, like 0hd, 10hd, and even exotic ones, considerably saving on tape length. Dynamic glider tapes might still win for straight-up speed, but here, single channel is no longer the clear winner, or practically any winner at all, opening up the door for something far more interesting...
256PS!
Please use our method unless it’s actually bad, or use a triangular memory region. We wouldn’t be happy if another boring old straight line comes out, and you guys all know what happens if we’re not happy...
Half-baked Demoni Models (that don’t involve half-bakeries)
Download | ???
Please adjust my calculations to correct for the Doppler effect.