The most recent Demonoid design had 100 still lifes per replicator unit, or 200 still lifes in all (not counting the self-destruct circuits, which never quite got finished to my satisfaction.)
The construction of the linear replicator (henceforth known as the "GoL propagator", by the way, to avoid terminology wars about the precise definition of "replicator") showed that it doesn't take very much Spartan Herschel circuitry to build a working construction arm. It should now be possible to make a diagonal Geminoid fly with only 44 still lifes per constructor unit, or 88sL in all:
Code: Select all
x = 1429, y = 1374, rule = LifeHistory 132.A$130.3A$129.A$129.2A8.2A$139.A$137.A.A$137.2A12.2A$150.A.A$150.A 11.A$127.2A20.2A9.3A$127.2A30.A$159.2A10.2A$108.2A61.A$108.2A59.A.A$ 169.2A$146.2A$146.2A41.A$179.2A6.3A$179.2A5.A$186.2A3$185.2A$180.2A3. 2A$155.2A23.2A$156.A$153.3A$118.2A33.A$118.2A22.2A47.A7.A$142.A.A44. 3A5.3A11.A$144.A43.A7.A14.3A$144.2A42.2A6.2A16.A14.A$213.2A12.3A$226. A$226.2A3$225.2A$206.2A17.2A$206.2A6$209.2A$210.A$207.3A$207.A$212.2A $213.A$210.3A$210.A8$222.2A$222.2A12$223.2A$223.A.A$225.A$225.2A2$8. 2A8.A$8.2A6.3A$15.A$2A13.2A$.A$.A.A$2.2A$7.2A204.2A$7.2A204.2A4$4.A$ 3.A.A$3.2A219.2A$203.2A19.A$204.A17.A.A$12.2A190.A.A15.2A$12.2A191.2A 4.A$210.A.A$210.A.A$211.A10.2A$222.A.A$224.A$224.2A$209.2A$210.A$207. 3A$207.A$223.3A$223.A$224.A70$98.2A$99.A$99.A.A$100.2A2$115.2A$115.2A 2$106.2A$98.2A6.2A$98.2A3$111.A$110.A.A$103.2A6.2A$104.A$101.3A$101.A 35$348.3A$348.A$349.A123$473.3A$473.A$474.A187$663.2A$663.A.A$663.A 596$1396.2A$1396.A$1394.A.A$1352.A37.2A2.2A$1326.A9.A15.3A35.2A$1326. 3A5.3A18.A$1329.A3.A20.2A11.2A$1328.2A3.2A32.2A8$1342.2A52.2A$1342.2A 34.2A16.A.A$1330.2A45.A.A18.A$1329.A2.A44.A20.2A$1324.2A4.2A44.2A4.2A $1323.A.A55.A.A$1323.A57.A$1322.2A56.2A7.2A$1332.2A55.2A$1332.A$1333. 3A$1335.A4$1399.2A$1399.A$1397.A.A$1397.2A5$1399.2A$1399.A$1397.A.A 11.2A$1397.2A12.A$1409.A.A$1405.2A2.2A$1405.2A4$1404.2A$1404.2A4.2A$ 1410.2A7$1417.A$1415.3A$1414.A$1414.2A7$1419.2A$1419.A$1417.A.A$1417. 2A3$1403.2A$1402.A.A$1402.A$1401.2A$1421.2A$1422.A$1419.3A$1419.A2$ 1412.2A$1412.2A$1397.A$1397.3A$1400.A$1399.2A2$1425.A$1423.3A$1422.A$ 1422.2A5$1427.2A$1427.A$1425.A.A$1425.2A$1418.2A$1418.2A8$1400.2A$ 1400.2A2$1243.2A$1243.2A3$1234.2A$1234.A.A$1235.A$1415.2A$1415.2A$ 1240.2A$1240.2A2$1233.2A$1232.A.A$1232.A$1231.2A13.2A$1246.A$1239.2A 6.3A$1239.2A8.A80$1349.2A$1349.A$1347.A.A$1347.2A2$1332.2A$1332.2A3$ 1349.2A$1342.2A5.2A$1342.2A2$1337.A$1336.A.A$1336.2A6.2A$1344.A$1345. 3A$1347.A!
Assume for the moment that the previous construction unit (southwest of the green one) has already been shot down and destroyed completely -- there's nothing left of it. Then the active constructor (green) has to
- build the main body of the new U.C. (white);
- build a series of two freeze-dried slow salvos to the north (yellow and green arrows);
- go inactive after sending out a final glider;
- have the glider trigger a previously-built copy of the first freeze-dried salvo (green rectangle) to shoot down the now-inactive U.C., leaving only two target ash objects (orange dots to the northeast);
- have the previous reaction trigger an older copy of the second freeze-dried salvo (yellow rectangle) which builds the two white semi-Snarks, using the older ash objects as targets (orange dots to the southwest).
- minor cleanup detail: send a few gliders from somewhere to shoot down the green semi-Snarks in the southwest.
With any luck, the first freeze-dried salvo can be very short -- hopefully just one glider! I've left room for a long destruction salvo in the blueprint, and that method would certainly work, but I think it would be much less efficient.
I believe there have to be two stages -- destruction of the main body of the U.C., then construction of semi-Snarks on the same site in a later cycle -- because the semi-Snarks at each stage have to shoot zillions of gliders right through the middle of the next semi-Snark locations. There's no way around this that I can see, without greatly increasing the total amount of circuitry needed.
Now, this may look like a crazy amount of multi-stage self-construction trickery... and it probably is. But here's why this is a really good practice pattern for an eventual quadratic-growth replicator construction:
- 1) There's an option to use self-destruct circuitry on the main body of the U.C. -- I still want to see a good single-glider-triggered self-destruct circuit in operation.
- 1b) The main body is only 31 still lifes, all in an easy line with plenty of space around it, so with any luck there will be a little competition here on the forums, to use the Seeds of Destruction Game to find the most efficient self-destruct circuitry. That would save me some work...
- 2b) With any luck, freeze-dried salvos for constructing just those teeny little semi-Snarks won't be too ridiculously huge -- see below.
-- Okay. So before we go any further with this design, does anyone see a radically simpler way to design a Demonoid? I'm definitely open for suggestions...!
I'm guessing that each semi-Snark will need about 60 gliders to construct -- maybe a few more if we include some well-placed extra junk to help with the destruction stage. We can probably freeze-dry a 60-glider salvo at a cost of three stable objects per glider.
So that means we're building a 60*3*2 = ~360-object constellation just to put the dozen still lifes for the semi-Snarks in the right place at the right time -- plus another fifty or sixty for the actual U.C.-plus-self-destruct. That's definitely doable, but it adds up to a much bigger construction project than the GoL propagator. The original Gemini spaceship did more work than that, but it had two arms to do the work with...
Anyone see a way to reduce the total construction burden for this Demonoid design? Maybe one of the original Demonoid blueprints will be cheaper after all -- they don't have the overlapping semi-Snark problem, so the total cost might be somewhere around 200sL instead of 420sL.
Sadly, there would be no excuse to get some experience with freeze-dried salvos... but to make up for that, we could use elbow-op recipes that included singleton gliders. Especially at 10hd, those seem to be significantly more efficient -- and the old Demonoids could handle input data four times faster, on top of that.