HartmutHolzwart wrote:Since Dave Greene's last post in December, there has been no report on further progress. Is this just lack of time or rather because of upcoming problems in the approach?
Mostly lack of time. I have one more month of work left at my current job -- lots of loose ends to wrap up, so unfortunately I'm temporarily having to avoid Life projects that require many consecutive late nights.
The next major subproject is to "seed" the new smaller Geminoid design with a minimum number of extra still lifes, out of the way of the actual circuitry, so that a single glider can trigger the clean self-destruction of the entire pattern. I hope to be able to take a couple of weeks in July to work on a new search utility to do the seeding.
The utility will be interactive instead of exhaustive, displaying several possible partial solutions for each round of searching, and letting a user pick whichever option looks best to use as a base for the next search. (Now that Golly's scripting system can capture mouse clicks as well as keyboard input, this should be fun to write.) With any luck it will be possible to develop some intuition about what "looks best". Maybe if enough people try out the utility, we can get competitions going to produce better and better solutions for various Geminoid subpatterns...!
Anyway, it's an interesting problem. Clearly the search space is plenty big enough that there's always going to be a way to seed a pattern for self-destruction. The trick is in the optimization: how can we quickly get close to the best solution, without exhaustively searching a Universe-sized space?
HartmutHolzwart wrote:I think the idea of getting read of a specific (de-)constuction arm for destroying the old engine is very ambitious.
It's actually quite easy to come up with an upper-bound self-destruct seeding for any given pattern -- using only blocks, if necessary.
I'm sticking with a seeded-circuitry design instead of a destructor arm simply because A) it allows the new Geminoid to be much smaller and simpler, and B) I really want to see this kind of self-destruct sequence in operation! It reminds me somehow of setting up long chains of dominoes and knocking them over, which I used to do a lot when I was growing up.
HartmutHolzwart wrote:Wouldn't it be a sensible intermediate step to copy the original recipe based on new technology? Even this more modest goal seems quite a lot of work to me, but absolutely worth the effort.
Oddly enough, by the time you've replaced all the technology that can be improved in the old Gemini, there's literally nothing left -- it ends up being a complete redesign. There are some signal-crossing problems that have to be solved all over again for the new Geminoid, especially if there's a separate destructor arm. I don't want to bother dealing with those synchronization issues, because it seems so easy (relatively speaking) to avoid them entirely.
But if anyone wants to tackle a three-arm Geminoid that uses one of Paul Chapman's new minimal two-lane salvo sets, well, more power to 'em! I'd be happy to assist with some of the circuitry design, but will stop short of writing the actual compiler to handle synch issues in the recipe glider stream(s).
HartmutHolzwart wrote:If this is not challenging enough, how about using p30 or p46 technology for the deconstruction arm. The design for that is much more compact (and thus should cheaper and faster to build) and also could fire destructing gliders at higher frequency.
The short answer here is that I don't think adding p30 or p46 to the mix will ultimately make a cheaper or faster-to-build Geminoid. What you gain in the compactness of the reflectors and splitters, you lose again in synchronizing all those oscillating components during the construction process.
I suppose you *might* get something a little smaller once all the design and construction problems were solved, but there's quite a bit of new research to do there. By contrast, the stable single-channel Geminoid is pretty close to being ready to go, except for figuring out exactly what glider stream to feed it. (!) Proving the viability and compactness of a p30/p46 Geminoid may ultimately require building a working prototype -- if it's actually smaller than whatever I come up with, well, then I was wrong...
The
"new way of constructing" thread showed off some very interesting p30 construction technology -- but it wasn't exactly fast or compact. But maybe I just don't have a clear picture of the modification you're suggesting here.
Keep the cheer,
Dave Greene
