Serizawa - Linear Self Replicator.

For discussion of other cellular automata.
User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 17th, 2014, 6:20 pm

@oblique. Yes. You see usually regular salvos recipes allow both glider colors. But in the approach borrowed from Serizawa "opposite direction collisions" which work nicely in orthogonal case, in Life using gliders you could only create one colored gliders. One approach is to use "two units", one for white gliders other for black. The other approach is to find some recipes for single colored salvos. The third is also possible, and it's to find recipes for "opposite collinear collisions", like the one I posted above, that create both black and white gliders.

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 18th, 2014, 6:12 am

Here is a recipe I found, with 3 pairs for white, 4 pairs for black collinear collisions (on the same tracks of course).

Code: Select all

x = 151, y = 228, rule = LifeHistory
149.B$148.3B$147.ABAB$147.2AB$148.A31$114.B$113.3B$112.A3B$112.ABA$
112.2A11$99.B$98.3B$97.ABAB$97.2AB$98.A26$69.B$68.BAB$67.2A2B$68.2A2$
60.3A$62.A$61.A18$40.3A$42.A$41.A18$20.3A$22.A$21.A9$113.B$112.A2B$
111.A3B$111.3A6$3A$2.A$.A6$96.B$95.BAB$94.2A2B$95.2A24$69.B$68.BAB$
67.2A2B$68.2A2$60.3A$62.A$61.A18$40.3A$42.A$41.A18$20.3A$22.A$21.A!
So I guess it's preferable to "singled colored" recipes, or "two interpreters" each for different color.

Another question that bothers me, is there a good collection of recipes for glider shoot from slow salve? everything I could find is of construction nature - block moves etc.

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 18th, 2014, 7:20 am

simsim314 wrote:Another question that bothers me, is there a good collection of recipes for glider shoot from slow salv[o]? everything I could find is of construction nature - block moves etc.
Well, there is a minimal toolkit available -- and if you remove the last glider from the bottom center block push, that could also be a way to split off a new "hand" block from an elbow, for recursive recipes that can reach around multiple corners.

I'm pretty sure there's no really well organized collection of slow-salvo OTT (one-time turner) seed recipes. There are uncounted millions of glider-turning small constellations out there, and it would be really nice to get a good pile of them all categorized by output lane and timing, along the lines of the blockish-and-blockic-seeds.rle pattern in Golly's Life/Syntheses folder.

Those Blockic seeds are in fact fairly inexpensive to build with slow salvos -- under nine gliders each, I believe -- but I don't know the recipes! They came from a data dump from Paul Chapman's Glue project, and the full recipes were not included. (Either that or I didn't save them, but I don't think I had that option.)

A good OTT collection including the delay information would allow not only recursive slow-salvo constructions, but also much more efficient seeds for complex glider constructions, along the lines of the loafer seed. There's really no reason to use just blocks in that recipe -- it just made it a lot easier to write the compiler script!

If OTT constellations could be built all at once instead of one object at a time, with the most efficient recipes -- which would be different for different orientations of OTT -- it would cut the loafer-seed recipe down to a small fraction of its current size.

Besides that, there are various other known single-object and small-constellation turners, so we could just build any of those and then hit it with the correct glider.

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 18th, 2014, 7:46 am

oblique wrote:I'm not quite getting the search space you are describing here. Do you mean: all gliders in one recipe must have the same color? Like in skiping every other "lane"? What are the other parameters of the search?
I'm still curious about whether monochromatic slow salvos are universal, even though simsim314 has now shown that single-color salvos aren't really necessary for his method.

To prove universality for this toolkit, we'd basically need to be able to move a block around freely, and generate both colors of 90-degree gliders. Diagonal block pulls can be monochromatic, and each (2,1) pull changes the color of the block:

Code: Select all

x = 17, y = 23, rule = B3/S23
10b3o$12bo2b2o$11bo3b2o8$3o7b3o$2bo9bo$bo9bo8$3o$2bo$bo!
We also need to be able to push a block away from the glider source, and place it anywhere we want. It turns out there are 8313 monochromatic recipes in my database of 1.8 million -- see attached. Here's the most efficient (-12,-10) block move, for example:

Code: Select all

#C 8-glider m-12,-10:E-9 E-13 E-13 O-13 E-9 E-5 O-9 O-5
x = 235, y = 238, rule = B3/S23
2o$2o4$3o$o$bo28$28b3o$28bo$29bo28$58b3o$58bo$59bo28$89b2o$88b2o$90bo
28$120b3o$120bo$121bo28$152b3o$152bo$153bo28$181b2o$180b2o$182bo48$
233b2o$232b2o$234bo!
As usual, the recipes are given in quarter-diagonals, so only odd lane numbers are possible for gliders. The main advantage is that any recipe can be trivially negated to get its mirror-image.

That just leaves the problem of finding monochromatic one-time-turner recipes for both colors of output gliders. At this point it's fairly clear that such things exist, so it's just a question of how organized we want to be in searching for them...!
Attachments
monochromatic-all.txt
All known monochromatic recipes for ash constellations
(388.7 KiB) Downloaded 776 times
monochromatic-slow-salvo-block-moves.txt
8313 single-color slow-salvo recipes from the P2 block-move table
(359.22 KiB) Downloaded 832 times

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 18th, 2014, 8:28 am

simsim314 wrote:Here is a recipe for a 3 collinear glider pairs for "slow" shoot. I also "proved" just by brute force that two collinear glider pairs always leave some debris.
Did you look at all seven possible sets of opposing glider lanes? The two results you posted have different offsets, 5hd and 4hd respectively, so I'm guessing you did. I'm just a little surprised that there isn't a 2gp glider output somewhere in that search space.
simsim314 wrote:Let's do some thinking: assuming we'll find some compact "two-glider input" duplicator with simple interpreter (one comes to my mind is just two silvers with two semi Snarks, allowing any distance we choose).
It will also be worth revisiting the idea of using tandem gliders instead of single gliders to carry information. Herschel transceivers are not much bigger than Silver reflectors, and can process signals about four times as fast.
simsim314 wrote:Also as a thought it might be possible to create some part of the new copy using direct shoots from the salvo itself, which might make it even faster than current Geminoid. Of course tweaking the Geometry is necessary, placing most of the "mass" of the new copy in the "shoot range" of the salvo.
Yes, this was always a big problem with the geometry of the one-arm Geminoids, especially for the "Demonoid" diagonal spaceship. The simplest Demonoid designs all needed at least one semi-Snark right on the construction lane... that construction could be done, theoretically, but only in the most ridiculously convoluted way (build a "meteor shower seed" and then trigger it at the very end of the construction to produce a construction salvo from a new direction).

With this new method, I think there's no particular problem with building objects on the construction lane, as long as they're built last. And the construction lane can be quite wide, though of course not quite as wide as the replicator unit! Some variant of this idea might even set a new record for the smallest Life universal constructor, which is currently somewhere around 44.5 still lifes in about a 200x200 bounding box.

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 18th, 2014, 2:41 pm

dvgrn wrote:Did you look at all seven possible sets of opposing glider lanes?
Well I'm bad at calculations, so I just placed all the combinations including a bit of those that "out of range". Than I took all the opposite collisions that were "reasonably small", i.e. I didn't take the pi or H debris. But yes I took everything "reasonable". Then I just search inside 120 possible 2-nd collisions, going +/- 15 on all four "relative glider states". 15 was enough because most of the "reasonable" cases that did not explode were treating the 15 gap as two separate inputs. Checking all the "nice" cases again, and there was always some leftovers.

So yes it's not "guaranteed" because I've checked it "by hand" (wrote a small script for all the possible collisions and then just looked at the results), so it's possible that I missed something, but it's pretty improbable, I looked at a lot of cases, there were debris everywhere :) .

More than that I can guarantee that this is the only case that shoots black and white glider after 2 glider shoots, and that the case with 4 shoots can't be improved to 3.

I'm not absolutely certain that this is the best recipe, but I'm quite sure it is, the only case I could miss is some glider shoot at "second shoot" that can be cleaned up at third, and some "miracle" at third that shoots and cleaned a glider after some "mysterious" second shoot. Although it's possibility I haven't check, it's really improbable on my opinion.

Anyway I thought for a moment that you meant my new recipe is not of the same track (I simply misread), but just for the clarity of things (and because I've already prepared it), here is a demonstration of the two recipes align properly together having their gliders output one cell apart:

Code: Select all

x = 401, y = 398, rule = LifeHistory
399.B$398.3B$397.ABAB$397.2AB$398.A31$364.B$363.3B$362.A3B$362.ABA$
362.2A11$349.B$348.3B$347.ABAB$347.2AB$348.A26$319.B$318.BAB$317.2A2B
$318.2A73$243.B$242.A2B$241.A3B$241.3A14$226.B$225.BAB$224.2A2B$225.
2A24$199.B$198.BAB$197.2A2B$198.2A2$190.3A$192.A$191.A18$170.3A$172.A
$171.A18$150.3A$152.A$151.A88$60.3A$62.A$61.A18$40.3A$42.A$41.A18$20.
3A$22.A$21.A18$3A$2.A$.A!
dvgrn wrote:...using tandem gliders
I guess you mean few glider tracks?
dvgrn wrote:Herschel transceivers are not much bigger than Silver reflectors
Hmm...do you mean there is a faster way to convert glider into Hershel?
dvgrn wrote: Some variant of this idea might even set a new record for the smallest Life universal constructor
I've been playing with it for a while, and annoyingly the large glider space between two glider shoots, caused by silver, makes it a bit harder than what I was expecting. So probably it could be twisted to make it smaller than the current UC, but it's definitely not so simple as it might seem. I was trying to create it with similar constrains (no Snarks), adding Snarks could improve it drastically, but the original UC as well.
Last edited by simsim314 on April 18th, 2014, 2:55 pm, edited 1 time in total.

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 18th, 2014, 2:52 pm

dvgrn wrote:Some variant of this idea might even set a new record for the smallest Life universal constructor, which is currently somewhere around 44.5 still lifes in about a 200x200 bounding box.
Early results are very promising. Here's a draft 39-still-life design in a 124x133 bounding box that might be programmable for universal construction. Might need a longer delay on the northwest end to give a wider range of allowable collision points, but that's a minor adjustment -- there are a lot of possible Herschel-to-glider converters to try there.

Code: Select all

#C 124x133 39sL replicator unit, possibly?
x = 257, y = 267, rule = B3/S23
55b2o$54bo2bo$55b2o26$52b2o$52b2o7b2o$61bo$59bobo$59b2o4b2o$43b2o20bo$
44bo18bobo$44bobo16b2o$45b2o$83bo$73b2o6b3o$73b2o5bo$12b2o66b2o$12b2o
2$79b2o$3b2o69b2o3b2o$3bobo68b2o$4bo$51b2o$47b2o2b2o$9b2o35bobo36bo7bo
$9b2o35bo36b3o5b3o11bo$45b2o35bo7bo14b3o$2b2o78b2o6b2o16bo14bo$bobo
103b2o12b3o$bo118bo$2o13b2o103b2o$15bo$8b2o6b3o$8b2o8bo100b2o$100b2o
17b2o$100b2o6$103b2o$104bo$101b3o$101bo$106b2o$107bo$104b3o$104bo8$
116b2o$116b2o12$117b2o$117bobo$119bo$119b2o9$107b2o$107b2o6$118b2o$78b
2o17b2o19bo$79bo18bo17bobo$79bobo16bobo15b2o$80b2o17b2o4bo$104bobo$95b
2o7bobo$95b2o8bo10b2o$116bobo$118bo$78b2o38b2o$78b2o5b2o16b2o$85b2o17b
o$101b3o$91bo9bo$90bobo$83b2o6b2o$84bo$81b3o$81bo7$129b2o$129bobo$129b
o123$254b2o$254bobo$254bo!
EDIT: The orientation of this construction arm wouldn't work very well for making a Geminoid spaceship -- theoretically maybe you could bend the arm around an extra elbow, but that's kind of a ridiculous idea, and there would be signal-crossing problems in the southeast. A slight variation of the above would work just fine for a spiral-growth pattern, though, and might also be useful for a 2D quadratic-growth Geminoid.

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 18th, 2014, 3:39 pm

simsim314 wrote:
dvgrn wrote:...using tandem gliders
I guess you mean few glider tracks?
dvgrn wrote:Herschel transceivers are not much bigger than Silver reflectors
Hmm...do you mean there is a faster way to convert glider into Hershel?
Yes. There are two basic types of Herschel transceivers, which accept "tandem gliders" -- gliders on parallel lanes, at 2, 5, or 6 cells apart (for Paul Callahan's reflector) or 4 cells apart (for Guam's recent transparent-transmitter version). It's best to think of a tandem glider as a Herschel signal in a suspiciously glider-like form. Tandem gliders can follow each other at 117 ticks, if I recall correctly, as opposed to 497 or 575 for Silver glider-to-Herschel converters.

It's not entirely clear that using transceivers will improve anything, though. What you gain in the recovery time you tend to lose in the signal splitters -- it's hard to produce a new tandem glider signal without killing a Herschel to do it, and most transmitters are not easily constructible anyway.

Hold that thought, though -- I just had a new idea that needs investigating...

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 18th, 2014, 3:53 pm

dvgrn wrote:I just had a new idea that needs investigating..
Hmmm... meanwhile here is an "old one" in action :)

Code: Select all

x = 1997, y = 2008, rule = LifeHistory
1940.2A$1939.A2.A$1940.2A26$1943.2A$1934.2A7.2A$1935.A$1935.A.A$1930.
2A4.2A$1931.A20.2A$1931.A.A18.A$1932.2A16.A.A$1950.2A$1913.A$1913.3A
6.2A$1916.A5.2A$1915.2A66.2A$1983.2A2$1916.2A$1916.2A3.2A69.2A$1921.
2A68.A.A$1992.A$1944.2A$1944.2A2.2A$1903.A7.A36.A.A35.2A$1891.A11.3A
5.3A36.A35.2A$1889.3A14.A7.A35.2A$1873.A14.A16.2A6.2A78.2A$1873.3A12.
2A103.A.A$1876.A118.A$1875.2A103.2A13.2A$1981.A$1978.3A6.2A$1876.2A
100.A8.2A$1876.2A17.2A$1895.2A6$1892.2A$1892.A$1893.3A$1895.A$1889.2A
$1889.A$1890.3A$1892.A8$1879.2A$1879.2A12$1878.2A$1877.A.A$1877.A$
1876.2A9$1888.2A$1888.2A6$1877.2A$1878.A19.2A17.2A$1878.A.A17.A18.A$
1879.2A15.A.A16.A.A$1891.A4.2A17.2A$1890.A.A$1890.A.A7.2A$1879.2A10.A
8.2A$1878.A.A$1878.A$1877.2A38.2A$1892.2A16.2A5.2A$1892.A17.2A$1893.
3A$1895.A9.A$1904.A.A$1904.2A6.2A$1912.A$1913.3A$1915.A7$1866.3A$
1868.A$1867.A142$1722.A$1721.B2A$1721.A.A$1722.B143$1576.3A$1578.A$
1577.A135$1439.A$1438.B2A$1438.A.A$1439.B150$1286.3A$1288.A$1287.A
139$1145.3A$1147.A$1146.A147$996.3A$998.A$997.A142$852.A$851.B2A$851.
A.A$852.B143$706.3A$708.A$707.A133$571.2A$570.2B2A$570.BAB$571.B152$
416.3A$418.A$417.A138$276.2A$275.ABA$275.B.A$276.B147$126.3A$128.A$
127.A123$.2A$2B2A$BAB$.B!
I think it's "ugly" that it uses so little space for shooting range. Anyway it has around 15-20 cells width, which supposed to be enough for construction arm. The next step is to import the construction arm recipes. One detail about them: we don't actually need an INC and DEC commands. As I mentioned we only need some subset of them that allows any INC and DEC we like. I also think that the current approach of sending glider is kinda "weird" , I would go for sending glider from some debris and come back into block, instead of splitting the block and then use the debris. It's just seems to be a slowing factor.

Well new UC recipes are on the way :)

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 18th, 2014, 6:40 pm

I've "proved" universality of this constructor. Using trial and error I've found the bounds of both white and black glider shoots. Using a little bit modified recipes for arm movements it's easy to see that the bounds of the UC shoots are much wider than the trace of the lowest and the highest gliders in the recipes. To shoot a different color glider, one should move the block by (2,1) and use the same recipe.

Code: Select all

x = 4215, y = 4321, rule = LifeHistory
4158.2A3.3A$4158.2A3.A$4164.A$4161.2D$4161.2D8$4165.3A$4165.A$4166.A
6$4138.2A$4138.2A3$4142.2D2.3A$4142.2D2.A$4147.A13$4117.2D10.2F$4117.
D2A10.F23.3A6.3A$4118.2A34.A8.A$4155.A8.A$4118.A$4117.2A$4117.A.A10$
4127.2A51.3A$4127.A.A50.A$4127.A13.2A38.A$4140.2A$4142.A4$4144.2A$
4143.2A$4135.3A7.A$4135.A$4136.A13.A$4149.2A$4149.A.A4$4193.3A$4193.A
$4194.A3$4198.3A$4198.A$4199.A5$4206.3A$4206.A$4207.A3$4212.3A$4212.A
$4213.A59$4102.2A$4101.A2.A$4102.2A26$4105.2A$4096.2A7.2A$4097.A$
4097.A.A$4092.2A4.2A$4093.A20.2A$4093.A.A18.A$4094.2A16.A.A$4112.2A$
4075.A$4075.3A6.2A$4078.A5.2A$4077.2A66.2A$4145.2A2$4078.2A$4078.2A3.
2A69.2A$4083.2A68.A.A$4154.A$4106.2A$4106.2A2.2A$4065.A7.A36.A.A35.2A
$4053.A11.3A5.3A36.A35.2A$4051.3A14.A7.A35.2A$4035.A14.A16.2A6.2A78.
2A$4035.3A12.2A103.A.A$4038.A118.A$4037.2A103.2A13.2A$4143.A$4140.3A
6.2A$4038.2A100.A8.2A$4038.2A17.2A$4057.2A6$4054.2A$4054.A$4055.3A$
4057.A$4051.2A$4051.A$4052.3A$4054.A8$4041.2A$4041.2A12$4040.2A$4039.
A.A$4039.A$4038.2A9$4050.2A$4050.2A6$4039.2A$4040.A19.2A17.2A$4040.A.
A17.A18.A$4041.2A15.A.A16.A.A$4053.A4.2A17.2A$4052.A.A$4052.A.A7.2A$
4041.2A10.A8.2A$4040.A.A$4040.A$4039.2A38.2A$4054.2A16.2A5.2A$4054.A
17.2A$4055.3A$4057.A9.A$4066.A.A$4066.2A6.2A$4074.A$4075.3A$4077.A7$
4028.3A$4030.A$4029.A166$3860.A$3859.B2A$3859.A.A$3860.B143$3714.3A$
3716.A$3715.A159$3553.A$3552.B2A$3552.A.A$3553.B150$3400.3A$3402.A$
3401.A163$3235.3A$3237.A$3236.A147$3086.3A$3088.A$3087.A130$2954.A$
2953.B2A$2953.A.A$2954.B143$2808.3A$2810.A$2809.A123$2683.A$2682.B2A$
2682.A.A$2683.B150$2530.3A$2532.A$2531.A127$2401.3A$2403.A$2402.A147$
2252.3A$2254.A$2253.A166$2084.A$2083.B2A$2083.A.A$2084.B143$1938.3A$
1940.A$1939.A157$1779.2A$1778.2B2A$1778.BAB$1779.B152$1624.3A$1626.A$
1625.A162$1460.2A$1459.ABA$1459.B.A$1460.B147$1310.3A$1312.A$1311.A
147$1161.2A$1160.2B2A$1160.BAB$1161.B162$996.3A$998.A$997.A142$852.A$
851.B2A$851.A.A$852.B143$706.3A$708.A$707.A133$571.2A$570.2B2A$570.BA
B$571.B152$416.3A$418.A$417.A138$276.2A$275.ABA$275.B.A$276.B147$126.
3A$128.A$127.A123$.2A$2B2A$BAB$.B!
It has DEC3 and INC1 operations. It's easy to see that this is enough for universal movement. Universal glider shoot I mentioned above, just notice the recipe returns the block to it's original diagonal with DEC4 operation (starting the recipe with (2,1) displacement will yield the same result, just in the end there would be one less (2,1) displacement shoot).

Now it would be nice to build something with it... just as an experiment. By the way do you have some "conventions" for salvo recipes? I saw that you have script for some "code", but I couldn't understand it, just saw it worked. I could do some "reverse engineering", just find gliders, recognize their locations. But if there is something out of the box, simple enough, I could definitely use it.
EDIT: I've found this one http://conwaylife.com/forums/viewtopic.php?p=8186#p8186 and it's definitely was understandable. The only question is what if I want to create some "pattern" is there a recipes for still life creation as well?
EDIT2: I've found this one http://conwaylife.com/forums/viewtopic.php?p=9590#p9590 but it's definetly a partial info, that does not include the location of each still life that was created.

I think we need some standard notions for still life, their orientation and locations (like we have for gliders). I guess you have some "standard" as a working prototype that you used in the Geminoid project, but I can't find any mention of it.

Anyway for me, there is enough info to play around with this new UC.

EDIT3: While I was thinking of it, due to the fact the current shooting range is actually considerably bigger than the optimized recipes range (which probably could be even more optimized as well), maybe it's possible to squeeze some extra still life and box-size optimization, cutting somewhere in the Hershel conduit...

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 19th, 2014, 1:06 pm

simsim314 wrote:I've "proved" universality of this constructor. Using trial and error I've found the bounds of both white and black glider shoots. Using a little bit modified recipes for arm movements it's easy to see that the bounds of the UC shoots are much wider than the trace of the lowest and the highest gliders in the recipes. To shoot a different color glider, one should move the block by (2,1) and use the same recipe.
Looks good! It's easy to find variations that widen the construction lanes -- it's just a matter of finding as efficient a way as possible to delay the right-hand glider. Here's a 41sL variant, for example -- gains 22 construction lanes:

Code: Select all

x = 1830, y = 1850, rule = LifeHistory
1784.A$1782.3A$1781.A$1781.2A6$1788.2A$1788.2A13$1772.2A$1772.2A20$
1770.2A$1761.2A7.2A$1762.A$1762.A.A$1757.2A4.2A$1758.A20.2A$1758.A.A
18.A36.2A$1759.2A16.A.A36.2A$1777.2A$1740.A$1740.3A6.2A74.2A$1743.A5.
2A73.A.A$1742.2A81.A3$1743.2A74.2A$1743.2A3.2A69.2A$1748.2A$1826.2A$
1771.2A53.A.A$1771.2A2.2A51.A$1730.A7.A36.A.A35.2A13.2A$1718.A11.3A5.
3A36.A36.A$1716.3A14.A7.A35.2A32.3A6.2A$1700.A14.A16.2A6.2A69.A8.2A$
1700.3A12.2A$1703.A$1702.2A3$1703.2A$1703.2A17.2A$1722.2A6$1719.2A$
1719.A$1720.3A$1722.A$1716.2A$1716.A$1717.3A$1719.A8$1706.2A$1706.2A
12$1705.2A$1704.A.A$1704.A$1703.2A9$1715.2A$1715.2A6$1704.2A$1705.A
19.2A17.2A$1705.A.A17.A18.A$1706.2A15.A.A16.A.A$1718.A4.2A17.2A$1717.
A.A$1717.A.A7.2A$1706.2A10.A8.2A$1705.A.A$1705.A$1704.2A38.2A$1719.2A
16.2A5.2A$1719.A17.2A$1720.3A$1722.A9.A$1731.A.A$1731.2A6.2A$1739.A$
1740.3A$1742.A7$1693.3A$1695.A$1694.A188$1503.2A$1502.A.A$1504.A144$
1357.3A$1359.A$1358.A181$1174.2A$1173.A.A$1175.A151$1021.3A$1023.A$
1022.A184$835.A$835.2A$834.A.A148$685.3A$687.A$686.A130$553.2A$552.A.
A$554.A144$407.3A$409.A$408.A123$282.2A$281.A.A$283.A151$129.3A$131.A
$130.A126$.A$.2A$A.A!
It would only take a few more still lifes to free up 40 or 50 more construction lanes, and that might be the really interesting model to build something with (because of the new potential for direct slow-salvo constructions, with no elbow needed.)
simsim314 wrote:EDIT3: While I was thinking of it, due to the fact the current shooting range is actually considerably bigger than the optimized recipes range (which probably could be even more optimized as well), maybe it's possible to squeeze some extra still life and box-size optimization, cutting somewhere in the Hershel conduit...
Yes, there's probably a universal replicator unit with fewer than 39 still lifes, at the cost of a narrower shooting range. The 39sL model has 37 available output lanes, I believe, where the elbow recipes you posted only seem to need a width of 14 (or maybe a little narrower, as you say). I'll dig up the true-minimal R.U. sometime or other, maybe, but that doesn't seem to be the most interesting line of investigation at the moment.

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 19th, 2014, 6:50 pm

Well I guess there is no choice other than to dive into Herschel conduits... it's unavoidable for effective life designs...

Anyway I'm conducting some experiment with small width salvos. I noticed that the current recipes didn't consider "glider range" as optimization parameter.

I'm currently investigating what is the smallest salvo width for "Universality". But Before that, I just want to have a "thin" recipe collection. I think it's more theoretical question, what is the smallest UC possible? Even if really ineffective. This question also has few meanings: is only Spartan components are permissible? or any still life? For computers guns and such there is no reason not to use the most advanced technology available, but for replicators it's better to use the Spartan toolkit.

The other question which is more "practical" is: what is the smallest "Full Range" opposite collision UC. For this one I really need the Herschel covered.

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 19th, 2014, 9:21 pm

simsim314 wrote:...I'm conducting some experiment with small width salvos. I noticed that the current recipes didn't consider "glider range" as optimization parameter... I'm currently investigating what is the smallest salvo width for "Universality". But Before that, I just want to have a "thin" recipe collection.
One of the things I tried early on, to help optimize Geminoid construction recipes, was to sort the block-move table by glider range. See attached.

The width is the first number on each line -- or actually it's width-minus-1, the difference between the highest and lowest lanes. The whole spec is {high-low}|{low}|{high}~{glidercost}m{X},{Y}:{recipe}. Worked for me -- feel free to adapt it for your purposes.
simsim314 wrote:I think it's more theoretical question, what is the smallest UC possible? Even if really ineffective. This question also has few meanings: is only Spartan components are permissible? or any still life? For computers guns and such there is no reason not to use the most advanced technology available, but for replicators it's better to use the Spartan toolkit.
I've generally defined "smallest" as "smallest number of Spartan still lifes", with only minor attention paid to the bounding box. Theoretically it's perfectly okay to include non-Spartan components if they really improve functionality tremendously.

But I've been focused primarily on unidirectional slow-salvo constructions, and it takes a lot of gliders to build larger objects with slow salvos. A single Snark reflector would probably be about as expensive as a dozen complete Silver reflectors, for example. It wouldn't be nearly as painful to build a Snark reflector with slow glider pairs, Gemini-style, so I'm hoping someone will take up that challenge soon.
simsim314 wrote:The other question which is more "practical" is: what is the smallest "Full Range" opposite collision UC. For this one I really need the Herschel covered.
By "Full Range" you mean that you can generate a glider on any workable lane between the two semi-Snarks? I'll figure that out eventually, next time I have an hour or two to spare -- probably not this weekend.
Attachments
blockmove2-sorted-by-width-then-cost.zip
sort isn't quite perfect (should sort by X and Y separately)
(16.99 MiB) Downloaded 854 times

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 20th, 2014, 7:00 am

dvgrn wrote: sort the block-move table by glider range. See attached.
First of all thank you for sharing this patterns, they are definitely useful and interesting.

But as universality concerned, what we need is movement + glider shoots (black+white).

So first of all I've found a way to diagonally move traffic light with 2-3 glider shoots, and 4 range by your notation (I didn't get how you calculate range, for me it's just natural to use glider level of freedom. So having only 3 levels of freedom or making a mark of 6 "blue horizontal cells" in LifeHistory), I managed to create all necessary movements.

Now the tricky part is to find thin glider shooters.

Code: Select all

x = 426, y = 182, rule = LifeHistory
2.3A297.3A2$A5.A293.A5.A$A5.A293.A5.A$A5.A293.A5.A2$2.3A297.3A53$65.A
299.A$64.2A298.2A$64.A.A297.A.A59$124.3A296.3A$124.A298.A$125.A298.A
57$183.A$182.2A$182.A.A!
dvgrn wrote:it's perfectly okay to include non-Spartan components if they really improve functionality tremendously
Well eventually it all comes to purpose and cost. Let's say you use them for replicator, so probably they will cost you more, or if you place them in glider gun they will cost you less (if optimal glider gun defined by bounding box). Or as you mentioned, the cost varies for different types of replicators. Anyway what's nice about Spartan components is they are minimal and probably irreducible (can't be replaced with even smaller set). So although replaceable, every Spartan design can go into any other design (maybe with cost), but non-Spartan design, could just be not appropriate for some purposes (like placing Snark in slow salvo based replicator).

That said if something can be generated quickly with slow salvo I can't find any good reason to stick to very restricted Spartan components (like the dead spark coil in H -> tandem glider component).

---

Regarding minimalist UC's we need some factor for how much ticks it takes to recover. Silver based replicators are very slow, although small. On the other hand tandem based replicators, could be bigger but also faster. So reasonable UC efficiency factor would be - the number of still life multiplied by recovery time.

---

Another though: if we will use this arm-less approach for UC's, and use two units so we have 90 degree instead of slow salvo, I don't think we have good recipes for the parts that "outside the bounds" except of slow salvo (X2 because we have two units). It's probable that placing "outside the bounds" can be optimized with "90 degree" two glider shoot. Kinda "90 degree-two-glider based slow salvo".

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 20th, 2014, 9:55 am

simsim314 wrote:Now it would be nice to build something with it... just as an experiment.
All you need is a script that takes an arbitrary lane list as input, and generates the glider pairs needed to build that lane list. Then you can feed in whatever recipe you like: a loafer, or a Snark (though nobody has done a slow-salvo recipe for one of those yet -- it would be an interesting little project)... working up to a recipe for a full Serizawa-style UC. Do you have a name for your new 180-degree-gliders design, by the way, or will it have to be just "Serizawa-style"?
simsim314 wrote:By the way do you have some "conventions" for salvo recipes? I saw that you have script for some "code", but I couldn't understand it, just saw it worked. I could do some "reverse engineering", just find gliders, recognize their locations. But if there is something out of the box, simple enough, I could definitely use it.
EDIT: I've found this one http://conwaylife.com/forums/viewtopic.php?p=8186#p8186 and it's definitely was understandable. The only question is what if I want to create some "pattern" is there a recipes for still life creation as well?

EDIT2: I've found this one http://conwaylife.com/forums/viewtopic.php?p=9590#p9590 but it's definetly a partial info, that does not include the location of each still life that was created.
Right. I haven't really organized everything properly yet -- just developed a system that worked for me, at least well enough to finish a replicator recipe.

Here's a short summary of what I have so far. First, you've already seen the big table of block moves, and the table of other object constructions made by progressively truncating the block-move table. I also have a table of simultaneous two-block constructions (since very often the last glider in a block move just deletes another block). It's sorted by the distance between blocks, so I can look up the shortest way to build two blocks at the same time. Similarly, I have a table of block-plus-object constructions, so I can build an object and have a leftover block somewhere to continue the slow-salvo construction.

So... starting with a block at (X0, Y0), I look through the relevant block-plus-object recipes and decide which one I want to construct next. That recipe will require a block at (X1, Y1), so I look up the cheapest way to move a block by (X1-X0, Y1-Y0), in the block-move table. All of these lookups are semi-automated with helper scripts, by the way, but not the kind of script that anyone else would have an easy time learning how to use!

There's also a stamp collection of the best slow-salvo eater recipes that anyone has come up with so far, since those are the biggest problem in this kind of construction. Those recipes all include the new location of the target block -- that's the "mX,Y" part of the recipe name -- so the salvo-combining script doesn't lose track of where the "hand" is. I should really have a similar stamp collection for all the other Spartan objects -- just haven't gotten around to it yet.

Anyway, then I append the block-move recipe, followed by the block-plus-object recipe, to the growing slow salvo recipe, using the combine-p2-recipes.py script, keeping track of where the new block ends up. Repeat this for all objects I want to build, and the slow salvo is complete.
simsim314 wrote:I think we need some standard notions for still life, their orientation and locations (like we have for gliders). I guess you have some "standard" as a working prototype that you used in the Geminoid project, but I can't find any mention of it.
Definitely this all needs to be standardized at some point. It may possibly be better to design a system from the ground up, rather than attempting to adapt my partial system. The main goal will be to come up with an "object compiler" that can start from an RLE pattern, or perhaps an ordered list of small objects, and reliably create a slow-salvo recipe that can build the full pattern.

I'm partway through writing a script along those lines, but the project is currently stalled at the "Blockic compiler" stage. That's the script I used to compile the now-outdated loafer-seed recipe.

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 20th, 2014, 1:21 pm

simsim314 wrote:So first of all I've found a way to diagonally move traffic light with 2-3 glider shoots, and 4 range by your notation (I didn't get how you calculate range, for me it's just natural to use glider level of freedom. So having only 3 levels of freedom or making a mark of 6 "blue horizontal cells" in LifeHistory), I managed to create all necessary movements.
Wow! INC and DEC control with a width of only three glider lanes seems pretty good. Then again, one of Paul Chapman's early searches found a universal toolkit with slow glider pairs following each other on a single lane, so I suppose I shouldn't be too surprised.
simsim314 wrote:Now the tricky part is to find thin glider shooters.
This is another problem that it shouldn't be too difficult to adapt oblique's sscs search program to look at: an exhaustive search of the slow-salvo tree at width 2, then width 3, then width 4... until a universal set of operations is found. I'm betting that width 4 will be enough, but width 5 and up will be progressively more efficient of course.
simsim314 wrote:Another though[t]: if we will use this arm-less approach for UC's, and use two units so we have 90 degree instead of slow salvo, I don't think we have good recipes for the parts that "outside the bounds" except of slow salvo (X2 because we have two units). It's probable that placing "outside the bounds" can be optimized with "90 degree" two glider shoot. Kinda "90 degree-two-glider based slow salvo".
Yes, it seems as if glider-pair slow salvos using two elbows would be correspondingly much more efficient than one-arm singleton glider salvos using one elbow. You'll still be able to collide two gliders at the construction site, which seems to cut down the total number of gliders needed by a factor of 5-25. (Tricky constructions like Snarks would get more like a 25-fold improvement, where Spartan objects like eaters are only 5-fold.)

Unrelated thought: It's possible to get 4-cell offsets for the 180-degree gliders, for only half of the small alternate Serizawa-type RUs that I'm working out timings for. The other half could have 3-cell or 5-cell offsets. Should I save those? i.e., is there an equivalent set of BLACK and WHITE glider outputs at some odd offset?

EDIT:Here's a sample 45sL replicator unit with a somewhat wider shooting range. Still looking for more efficient options...

Code: Select all

x = 250, y = 269, rule = B3/S23
218b2o$218bo$216bobo$216b2o7$211b2o4b2o$173b2o36b2o4b2o$172bobo$172bo$
171b2o$212b2o$212b2o2b2o$216bobo$218bo$218b2o7$204b2o$204bobo$206bo$
206b2o7$196b2o$187b2o7b2o$188bo$188bobo$183b2o4b2o45b2o$184bo20b2o29b
2o$184bobo18bo$185b2o16bobo$203b2o40b2o$166bo77bobo$166b3o6b2o68bo$
169bo5b2o$168b2o$239b2o$239b2o$169b2o$169b2o3b2o70b2o$174b2o70bobo$
248bo$197b2o34b2o13b2o$197b2o2b2o31bo$156bo7bo36bobo27b3o6b2o$144bo11b
3o5b3o36bo27bo8b2o$142b3o14bo7bo35b2o$126bo14bo16b2o6b2o$126b3o12b2o$
129bo$128b2o3$129b2o$129b2o17b2o$148b2o6$145b2o$145bo$146b3o$148bo$
142b2o$142bo$143b3o$145bo8$132b2o$132b2o12$131b2o$130bobo$130bo$129b2o
9$141b2o$141b2o6$130b2o$131bo19b2o17b2o$131bobo17bo18bo$132b2o15bobo
16bobo$144bo4b2o17b2o$143bobo$143bobo7b2o$132b2o10bo8b2o$131bobo$131bo
$130b2o38b2o$145b2o16b2o5b2o$145bo17b2o$146b3o$148bo9bo$157bobo$157b2o
6b2o$165bo$166b3o$168bo2$124b3o$126bo$125bo122$2o$b2o$o!

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 20th, 2014, 1:28 pm

dvgrn wrote:Do you have a name for your new 180-degree-gliders design
Well it's "my idea" (although it's pretty obvious idea, I just executed it in Serizawa and thought of "actually doing" the same in Life, so my idea starts and ends on "lets do it", because we all were thinking it at some point), anyway it's probably more your design than mine (at least the execution).

I thought we are talking about two designs. One is armless UC, the other is VSUC (very small universal constructor). The working prototype we have, is too big for VSUC (although giving the right credit as for now it's the best VSUC we got), and too small for armless UC. Anyway they both belong to UC's based on 180-degree-"opposite glider collision" (OGC?). Not sure about VSUC and OGC but armless UC sound good to me. The first creature that actually "happy" to be without arms :)

---

Anyway "back to business" I was thinking of replication design, that will "fit" into the armless creature fully. More precisely it itself will fit inside itself. Because the armless UC is thin and wide, If we orient it 90% it will fit completely inside it's construction line. So the "full" replication will actually consist of two steps, one 90% turn and flip (or something along those lines), then 90 degree turn and flip again back to the original orientation, that will make kinda "steps" procedure. Now the tricky part is the "wiring", which need to be able to feed the construction from above, below, and from one side and the other side. But this "extra wiring weight" probably worth the full salvo construction.
dvgrn wrote:but not the kind of script that anyone else would have an easy time learning how to use!
I also use kinda "hardcore" scripts easily getting to few thousand rows. Like one that I've created Serizawa replicator with (actually ALL the replicator code is inside, you can run it as is, and will compile the whole replicator out of the box. And although it's not my "best script ever" it's actually not so bad, having a lot of classes and reusable code inside it - I actually intend to reuse it soon for the next Serizawa replicator). Somehow it seems people having hard time (me as well) use someone else's code, pretty much preferring or "run it out of the box" or write one on their own.

As I'm concerned, I will be glad to write the script fine enough so that I myself will prefer to reuse it rather than rewriting all from scratch. If I reach such level of coding, for me it can be called "good coding".
dvgrn wrote:Those recipes all include the new location of the target block
Looks great! At least one is covered :)
dvgrn wrote: I should really have a similar stamp collection for all the other Spartan objects -- just haven't gotten around to it yet.
You said you have a db of all those recipes, you just don't have time to compile a query? Or is it all somewhat more "messy"?
dvgrn wrote: that can start from an RLE pattern, or perhaps an ordered list of small objects, and reliably create a slow-salvo recipe that can build the full pattern
Recently I've posted a script that takes current state in golly, and counts the still life. Probably it can be adapted to "recognize" the objects, it shouldn't be too hard. Then we need some "system" that will translate those objects into salvos.

Because of my approach designing mostly by code from the beginning, what I did in Serizawa, is together with placing objects in the script (although in Serizawa there are two main object not dozens), I added them to some data list, and then wrote a small "compiler" that translate the data into glider stream (commands for arm movements and glider shoots, currently just sorted by diagonals as you saw). Anyway I think in Serizawa as well it would be nice to write a "pattern recognition script". I probably do it later today. What is fun in Serizawa, is that each object can be easily recognized as "duplicator" or "reflector" so it would be cool to have Serizawa geometry, translated into "input output" manner, as I usually code it.
dvgrn wrote:..."Blockic compiler" stage...
I was writing a Blockic compiler myself when I wanted to create p60 based replicator, using blockic. It was real mess. By the way it's not such a bad idea, having a replicator based on p60 (Gemini style). It works much faster but of course makes only blocks, but blocks are enough for any blockic setup so it's achievable and p60 is cheap so there is no need in thousands blocks. Probably it would end up somewhere near Gemini performance. My current guess the fastest approach for replicator would be adapting the current Geminoid for two arms. It's already works with 4 gliders inputs, just need to feed it to the other arm, and it will probably be the fastest Geminoid (maybe use tandems as well, and although not sure, but it's possible that more streams will make it more effective despite the sync issues).

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 20th, 2014, 1:57 pm

simsim314 wrote:
dvgrn wrote:I should really have a similar stamp collection for all the other Spartan objects -- just haven't gotten around to it yet.
You said you have a db of all those recipes, you just don't have time to compile a query? Or is it all somewhat more "messy"?
It's the half-million recipes I mentioned earlier in this thread -- just a flat file, not something that anyone else would think of as a queryable database.

I have a good "recognizer" script that can sort out the various small still lifes and p2 oscillators, and find their locations relative to the initial block. It actually wouldn't take very long to create a smaller database with all the block+object recipes, and register the output blocks so their locations are known.

It will take a bit longer to come up with a good "stamp collection" for each type of Spartan still life. Many recipes aren't really very usable for slow-salvo constructions, just because the output object isn't on the edge of the construction envelope (the blue area in LifeHistory). So a lot of recipes can be used to build an object in isolation, but not near other already-constructed objects.

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 20th, 2014, 2:07 pm

dvgrn wrote: Then again, one of Paul Chapman's early searches found a universal toolkit with slow glider pairs following each other on a single lane...
Well this one has much more degrees of freedom. Think of it statistically, let's say glider will come out by random 1 out of 100 patterns. having another degree of freedom to play with, gives you square more possibilities than only one salvo.

This brings me to another point - blinkers give you twice as much possibilities as block, and traffic light is also the most probable object in life and it's wide, having more targets to aim at. But yes, I started from block, and wanted to search lanes of two glider, it soon enough brought me to traffic light. Started to play with it instead of block and found those "nice" INC and DEC operations. But annoyingly enough it didn't brought any nice glider shoots, and I conduct a pretty nice depth search currently at depth of around 15 with about 30K patterns (I forgot to write the depth, and didn't wrote save function, so now just don't want to kill it, after a run of couple of hours).
dvgrn wrote:. I'm betting that width 4 will be enough
My bet is 3 - but as I mentioned the depth is pretty large, making it mainly theoretical query and not practical (no one wants to shoot 20-30 gliders for single reflected glider). Also noticing that VSUC has a really big margin between white and black patterns, it's useful to check single colored minimal "width" for universality. So currently this search is even not for sake of VSUC it's just a theoretical question that I'm trying to answer. The more VSUCish question will be: what is the single color minimal width for universality? My guess is also 3.
dvgrn wrote:is there an equivalent set of BLACK and WHITE glider outputs at some odd offset?
Well if I understand you correctly, BLACK and WHITE recipes are totally different, it's just pure chance they both ended in the same lanes. So no, there is no "odd" offset for timing. But maybe I misunderstood your question...

Some detail that might be bothering you (I know it from Serizawa, and I'm not sure it's your issue just something that caught me unprepared at the time): two opposite gliders even in the same "inner state" can have actually "two different 180 degree collisions". look at it in one dimension. Say you have two object moving toward each other
1. ...>...<....
2. ....>.<.....
3 ......X......

Now this is the odd case, there is also an even case:

1. ...>..<...
2. ....><...
3. ....XX...

So moving the interaction "to the left" on OGC-RU, will require actually steps of two in timing.

Hope it helped :)
Last edited by simsim314 on April 20th, 2014, 2:38 pm, edited 1 time in total.

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 20th, 2014, 2:16 pm

dvgrn wrote:It will take a bit longer to come up with a good "stamp collection"
Thinking of it - maybe it's possible to see what recipes were actually chosen by the compiler, to construct the linear replicator, and take them into stamp collection. Just a thought (I might prepare such script in the future myself).

I wanted to use the same idea, to take recipes from Gemini two arm construction. It's also not as trivial as it seems, Gemini uses pretty advanced placing patterns that go to the edge very efficiently.

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 20th, 2014, 3:05 pm

dvgrn wrote:Here's a sample 45sL replicator
Looks better...

I think that armless replicator, if we want it to fully replicate itself with slow salvo alone should probably be a bit wider than it's height. I think the current model need to be just a bit lengthened.

Code: Select all

x = 363, y = 316, rule = LifeHistory
162.B$161.4B$159.6B$157.9B26.2A$156.13B16.3B4.A$155.16B14.5BA.A$155.
16B13.6B2A$155.17B5.B6.6B$154.25B2.B.6B$152.38B$151.39B$150.39B$149.
40B.2B$149.36B2A4B2A$147.2AB2.24B3.6B2A2B.B2A$146.A.AB2.24B5.B2.5B2.B
$146.A6.23B8.6B$145.2A6.23B10.4B$154.B2.20B8.B2A2B$158.20B8.2A.B2A$
160.2B.16B10.BA.A$163.3B.8B.4B12.A$168.7B2.4B11.2A$168.7B3.4B$169.6B
4.4B$169.7B4.4B$169.8B4.4B$170.8B4.4B$170.8B5.4B$169.6B2.B2A4.4B$169.
7B.BA.A4.4B$170.6B4.A5.4B$170.6B4.2A5.4B$170.6B12.4B$169.8B12.4B$168.
8B14.4B$168.9B14.4B$168.9B15.4B$167.10B16.4B$167.3B2A5B17.4B$161.2A3.
4B2A5B18.4B$162.A3.11B19.4B$162.A.A12B20.4B$157.2A4.2A2.8B23.4B8.2A$
158.A9.7B4.2A18.4B6.B2AB$158.A.AB7.6B4.A20.4B5.4B$159.2AB.3B3.6B.BA.A
21.4B5.2B$161.14B.B2A23.4B2.4B7.2A$140.A20.16B26.4B.5B5.A.A$140.3A6.
2A11.14B28.14B.A$143.A4.B2AB9.16B28.14B$142.2A4.3B8.18B28.14B$142.5B
2.B2.2B2.20B28.9B2A4B$144.31B28.10B2A5B$143.2A21B.7B28.18B$143.2A3B2A
16B2.6B27.4B2.11B.B2A$144.B.2B2A16B3.6B25.4B5.8B2.BA.A$146.19B6.4B24.
4B4.10B5.A$148.10B12.B2A2B23.4B5.2A5.B.B5.2A$148.9B14.2A.B2A20.4B7.A
4.3B$130.A7.A8.9B18.BA.A18.4B5.3A5.B2AB$118.A11.3A5.3A7.7B22.A17.4B6.
A8.2A$116.3A14.A7.A6.6B23.2A15.4B$100.A14.A16.2A6.2A7.6B38.4B5.B$100.
3A12.2A15.4B4.4B4.8B36.4B5.3B$103.A7.2B.3B3.7B.B5.3B5.3B2.8B36.4B6.4B
$102.2A6.5B5.13B.4B4.13B35.4B8.4B$102.5B2.46B34.4B10.4B$104.51B33.4B
12.4B$103.2A50B32.4B14.4B$103.2A17B2A32B30.4B16.4B$104.B.16B2A34B24.A
2.4B18.4B$106.23B3.2B2.22B23.BA5B20.4B$107.9B2.10B11.20B22.BA4B22.4B$
108.8B3.6B13.21B20.7B24.4B$109.7B4.3B14.4B.17B19.3A3B3A24.4B$106.11B
3.B15.4B.19B19.7B26.4B$105.12B2.2A14.4B2.21B19.BA3B27.4B$105.12B2.A
14.4B3.21B19.BA3B28.4B$105.11B4.3A10.4B6.20B18.BA2B30.4B$105.8B.4B4.A
9.4B6.20B18.4B32.4B$105.7B4.2A13.4B7.19B18.4B34.4B$105.7B4.A13.4B9.
18B17.4B36.4B$105.6B6.3A9.4B10.17B17.4B38.4B$104.7B8.A8.4B12.16B16.4B
40.4B$103.8B16.4B14.14B16.4B42.4B$102.8B16.4B15.13B16.4B44.4B$101.9B
15.4B15.11B18.4B46.4B$100.4B.6B13.4B17.9B18.4B48.4B$99.4B.7B12.4B20.
7B17.4B25.2A23.4B$98.4B2.6B12.4B23.4B17.4B27.A24.4B$97.4B3.8B4.B4.4B
23.4B17.4B28.A.A23.4B$96.4B5.B2A6B.4B3.2B23.4B17.4B30.2A24.4B$95.4B5.
2B2A17B21.4B17.4B58.4B$94.4B7.19B21.4B17.4B60.4B$93.4B9.18B20.4B17.4B
62.4B$92.4B13.13B21.4B17.4B64.4B$91.4B13.12B22.4B17.4B66.4B$90.4B14.
10B23.4B17.4B68.4B$89.4B14.11B22.4B17.4B36.2A4.2A26.4B$88.4B15.7B.2B
22.4B17.4B37.2A4.2A27.4B$87.4B15.11B21.4B17.4B74.4B$86.4B17.11B19.4B
17.4B76.4B$85.4B18.11B18.4B17.4B78.4B$84.4B19.11B17.4B17.4B46.2A32.4B
$83.4B18.2AB2.8B16.4B17.4B43.2A2.2A33.4B$82.4B18.A.AB3.7B15.4B17.4B
43.A.A38.4B$81.4B19.A6.7B14.4B17.4B44.A41.4B$80.4B19.2A7.6B13.4B17.4B
44.2A42.4B$79.4B29.7B11.4B17.4B90.4B$78.4B30.8B9.4B17.4B92.4B$77.4B
32.8B7.4B17.4B94.4B$76.4B33.9B5.4B17.4B96.4B$75.4B33.6B.4B3.4B17.4B
98.4B$74.4B34.7B.4B.4B17.4B100.4B$73.4B36.6B2.7B17.4B65.2A35.4B$72.4B
37.6B3.5B17.4B65.A.A36.4B$71.4B38.2B2AB4.5B16.4B66.A39.4B$70.4B38.3B
2A2B2.7B14.4B66.2A40.4B$69.4B39.7B.4B.4B12.4B110.4B$68.4B41.5B.4B3.4B
10.4B112.4B$67.4B42.9B5.4B8.4B114.4B$66.4B43.8B7.4B6.4B116.4B$65.4B
44.7B9.4B4.4B9.B108.4B$64.4B36.2A7.6B11.4B2.4B9.3B108.4B$63.4B38.A7.
7B5.2A4.8B5.2A3.4B69.2A37.4B$62.4B39.A.AB3.7B6.A6.7B5.A5.4B68.2A7.2A
29.4B$61.4B41.2AB.11B.BA.A7.7B.BA.A6.4B76.A31.4B$60.4B44.10BA2B.B2A6.
9B.B2A8.4B73.A.A32.4B$59.4B44.10BABA3B5.2B.11B11.4B25.2A45.2A4.2A28.
4B$58.4B46.9BABA3B4.2A13B12.4B24.2A29.2A20.A30.4B$57.4B45.2AB.8BA2B6.
2A13B13.4B55.A18.A.A31.4B$56.4B45.A.AB2.8B9.2B2.10B14.4B54.A.A16.2A
33.4B$55.4B46.A4.4B2.3B14.9B.2B12.4B12.2A40.2A52.4B$54.4B46.2A3.4B3.
5B12.11B2A12.4B11.A.A77.A16.4B$53.4B51.4B6.B2A12.4B2A3B.B2A13.4B11.A
68.2A6.3A17.4B$52.4B51.4B8.A13.4B2A4B.B15.4B79.2A5.A21.4B$51.4B51.4B
10.3A10.7B21.4B85.2A21.4B$50.4B51.4B13.A9.A.5B23.4B13.2A93.4B$49.4B
51.4B23.A.A3.4B22.4B12.2A94.4B$48.4B51.4B24.2A6.2A23.4B81.2A25.4B$47.
4B51.4B33.A25.4B75.2A3.2A26.4B$46.4B51.4B35.3A23.4B74.2A32.4B$45.4B
51.4B38.A24.4B108.4B$44.4B51.4B65.4B13.2A34.2A57.4B$43.4B51.4B67.4B
12.A31.2A2.2A58.4B$42.4B51.4B69.4B4.2A6.3A27.A.A36.A7.A18.4B$41.4B51.
4B71.4B3.2A8.A27.A36.3A5.3A11.A7.4B$40.4B51.4B73.4B39.2A35.A7.A14.3A
6.4B$39.4B51.4B75.4B75.2A6.2A16.A6.4B$38.4B51.4B77.4B99.2A7.4B$37.4B
51.4B79.4B108.4B$36.4B51.4B81.4B108.4B$35.4B51.4B83.4B108.4B$34.4B51.
4B85.4B108.4B$33.4B51.4B87.4B108.4B$32.4B51.4B89.4B86.2A20.4B$31.4B
51.4B91.4B85.2A21.4B$30.4B51.4B93.4B108.4B$29.4B51.4B95.4B108.4B$28.
4B51.4B97.4B108.4B$27.4B51.4B99.4B108.4B$26.4B51.4B101.4B108.4B$25.4B
51.4B103.4B82.2A24.4B$24.4B51.4B105.4B82.A25.4B$23.4B51.4B107.4B78.3A
27.4B$22.4B51.4B109.4B77.A30.4B$21.4B51.4B111.4B81.2A25.4B$20.4B51.4B
113.4B81.A26.4B$19.4B51.4B115.4B77.3A28.4B$18.4B51.4B117.4B76.A31.4B$
17.4B51.4B119.4B108.4B$16.4B51.4B121.4B108.4B$15.4B51.4B123.4B108.4B$
14.4B51.4B125.4B108.4B$13.4B51.4B127.4B108.4B$12.4B51.4B129.4B108.4B$
11.4B51.4B131.4B108.4B$10.4B51.4B133.4B80.2A26.4B$9.4B51.4B135.4B79.
2A27.4B$8.4B51.4B137.4B108.4B$7.4B51.4B139.4B108.4B$6.4B51.4B141.4B
108.4B$5.4B51.4B143.4B108.4B$4.4B51.4B145.4B108.4B$3.4B51.4B147.4B
108.4B$2.4B51.4B149.4B108.4B$.4B51.4B151.4B108.4B$4B51.4B153.4B108.4B
$3B51.4B155.4B108.4B$2B51.4B157.4B108.4B$B51.4B159.4B68.2A38.4B$51.4B
161.4B67.A.A38.4B$50.4B163.4B68.A39.4B$49.4B165.4B67.2A39.4B$48.4B
167.4B108.4B$47.4B169.4B108.4B$46.4B171.4B108.4B$45.4B173.4B108.4B$
44.4B175.4B108.4B$43.4B177.4B108.4B$42.4B179.4B108.4B$41.4B181.4B108.
4B$40.4B183.4B46.2A60.4B$39.4B185.4B45.2A61.4B$38.4B187.4B108.4B$37.
4B189.4B108.4B$36.4B191.4B108.4B$35.4B193.4B108.4B$34.4B195.4B108.4B$
33.4B197.4B50.2A56.4B$32.4B199.4B9.2A17.2A19.A58.4B$31.4B201.4B9.A18.
A17.A.A59.4B$30.4B203.4B8.A.A16.A.A15.2A61.4B$29.4B205.4B8.2A17.2A4.A
74.4B$28.4B207.4B31.A.A74.4B$27.4B209.4B21.2A7.A.A75.4B$26.4B211.4B
20.2A8.A10.2A65.4B$25.4B213.4B40.A.A65.4B$24.4B215.4B41.A66.4B$23.4B
217.4B40.2A66.4B$22.4B219.4B6.2A16.2A82.4B$21.4B221.4B5.2A17.A83.4B$
20.4B223.4B20.3A85.4B$19.4B225.4B9.A9.A88.3B$18.4B227.4B7.A.A98.2B$
17.4B229.4B7.2A99.B$16.4B231.4B$15.4B233.4B$14.4B235.4B$13.4B237.4B$
12.4B239.4B$11.4B241.4B$10.4B243.4B$9.4B245.4B$8.4B247.4B$7.4B249.4B$
6.4B251.4B$5.4B253.4B$4.4B255.4B$3.4B257.4B$2.4B259.4B$.4B261.4B$4B
263.4B$3B265.4B$2B267.4B$B269.4B$271.4B$272.4B$273.4B$274.4B$275.4B$
276.4B$277.4B$278.4B$279.4B$280.4B$281.4B$282.4B$283.4B$284.4B$285.4B
$286.4B$287.4B$288.4B$289.4B$290.4B$291.4B$292.4B$293.4B$294.4B$295.
4B$296.4B$297.4B$298.4B$299.4B$300.4B$301.4B$302.4B$303.4B$304.4B$
305.4B$306.4B$307.4B$308.4B$309.4B$310.4B$311.4B$312.4B$313.4B$314.4B
$315.4B$316.4B$317.4B$318.4B$319.4B$320.4B$321.4B$322.4B$323.4B$324.
4B$325.4B$326.4B$327.4B$328.4B$329.4B$330.4B$331.4B$332.4B$333.4B$
334.4B$335.4B$336.4B$337.4B$338.4B$339.4B!

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 20th, 2014, 3:42 pm

simsim314 wrote:
dvgrn wrote:It will take a bit longer to come up with a good "stamp collection"
...I wanted to use the same idea, to take recipes from Gemini two arm construction. It's also not as trivial as it seems, Gemini uses pretty advanced placing patterns that go to the edge very efficiently.
Calcyman (Adam P. Goucher) has already done a stamp collection for two-arm Gemini constructions --

Code: Select all

#C recent version of calcyman's SSL.mc, including tub-with-tail eaters
x = 4101, y = 373, rule = LifeHistory
1555.D$1555.3D$1558.D$1557.2D4$1820.D$1818.3D$1817.D$1817.2D2$83.2D
118.2D$83.2D118.2D229.D$433.D.D$433.D.D$434.D3431.2D$3866.2D2$4045.2D
$3865.D180.D$3864.D.D179.D.D$3865.D181.D.D$624.2D3236.3D183.D3.2D$
623.D2.D3235.D189.2D$624.2D6$4040.2C$63.3C156.3C3816.2C$65.C156.C
1298.3C2348.2C166.C25.2C$64.C158.C1299.C2321.2C25.C.C191.C.C$104.3C
74.3C1338.C2323.2C24.C193.C$104.C78.C3661.C$105.C76.C1402.2C$1585.C.C
$1585.C$451.3C1398.3C$451.C1400.C$452.C1400.C2170.2C$414.3C3436.3C
169.2C$416.C1372.2C2064.C168.C45.2C$415.C1372.C.C2063.C26.2C187.C.C$
1790.C2090.C.C186.C$3881.C2$649.3C$649.C623.2D$650.C623.D$597.3C674.D
.D$599.C675.2D2743.3C$598.C3301.3C$3900.C120.2C60.2C$3901.C180.2C$
804.2D3278.C$803.D.D$803.D$478.3C321.2D$478.C3353.2C$391.2C86.C3353.
2C$392.2C3438.C179.3C$391.C3421.3C198.C$3815.C197.C80.2C$1019.2D2793.
C102.2C175.C.C$1019.D.D2895.C.C174.C$1021.D2895.C$1021.2D$1616.2C$
1615.2C$1617.C581.2D$1252.3C944.D$1254.C945.3D$1253.C44.2C902.D$1297.
2C185.3C2314.C126.2C$1299.C186.C271.2C2041.2C124.2C$1485.C273.2C2039.
C.C126.C$1758.C$94.2D96.2D$94.2D96.2D2343.2D$2538.D$1889.3C643.3D$
782.3C1104.C645.D$784.C1105.C$783.C40.2C2359.D$823.2C1964.D394.D.D$
825.C1962.D.D394.D$2788.2D2$3539.2D$3539.D.D$1040.3C2497.2D128.2D$81.
2C122.2C833.C1890.2D736.D.D$80.C.C122.C.C233.D557.2C40.C1888.D.D736.
2D$82.C122.C234.D.D557.2C1929.D$104.3C74.3C256.D.D556.C$104.C78.C257.
D$105.C76.C3183.2D$3365.D2.D$3365.D2.D674.2D$3366.2D675.2D3.D$4047.D.
D$3858.D189.D.D$71.2C38.3C60.3C38.2C3639.3D191.D$72.2C140.2C3639.D
194.2D$71.C39.2C62.2C39.C407.2D2175.3C1050.D.D$623.D2.D2174.C1053.D$
624.2D2176.C$2770.3C$2772.C1080.2D$847.3C1921.C1081.2D206.2C$847.C
3212.2C$848.C3186.2C25.C$4034.C.C$1540.D2495.C$1540.3D620.C70.3C$761.
2C780.D619.2C69.C1631.2C$760.C.C779.2D618.C.C70.C712.3C915.C.C$762.C
2185.C917.C$975.3C1971.C891.2C$422.3C552.C1931.3C928.C.C$424.C551.C
858.D1075.C303.3C624.C183.3C$423.C1409.3D1074.C304.C812.C$459.3C1370.
D1383.C810.C36.2C$459.C1372.2D667.3C70.C565.3C358.C70.3C489.C.C$460.C
601.2C1439.C69.2C567.C358.2C69.C491.C$1062.C.C1437.C70.C.C565.C358.C.
C70.C58.C70.3C$1062.C2569.2C69.C$3631.C.C70.C165.2C$3870.C.C$3870.C$
598.3C251.3C2971.2C$600.C2727.C70.3C423.C.C$599.C252.2C2474.2C69.C
427.C$650.3C2674.C.C70.C$650.C3429.2C$651.C89.2C2749.3C90.3C425.2C64.
2C$740.C.C2232.3C516.C519.2C65.C$742.C2144.3C85.C517.C91.2C36.3C90.3C
294.C$2889.C86.C739.C$395.3C572.3C1915.C735.2C91.C166.2C$397.C3485.2C
$396.C86.2C486.2C2912.C$90.2D104.2D284.2C2750.3C$90.2D104.2D286.C
2749.C$1082.2C1040.3C1108.C583.C$1082.C.C1041.C601.3C118.3C967.2C$
1082.C1042.C604.C118.C968.C.C$1511.3C55.2C1158.C120.C$1513.C55.C.C
1541.2C875.3C$1512.C56.C1542.C.C877.C$3114.C876.C96.2C$4088.C.C$2272.
C1815.C$2271.2C338.3C$1805.2C55.3C406.C.C337.C1287.2C$1804.C.C55.C
749.C1286.C.C$1806.C56.C2035.C$3805.2C292.2C$3804.C.C291.2C$3806.C
293.C$76.3C15.2C96.2C15.3C$78.C15.C.C94.C.C15.C2255.C$77.C16.C98.C16.
C2254.2C$2464.C.C1515.2C$3983.2C$3982.C$1589.3C$1589.C$1590.C5$1784.
3C$1786.C$1785.C2$1484.2C$1485.2C$801.2D681.C$800.D.D$800.D$799.2D
2876.2D$3676.D.D$1890.2C1654.2D128.2D$104.2C76.2C1705.2C1655.D.D$104.
C.C74.C.C1707.C1655.2D$104.C78.C$1022.2D$1022.D.D$1024.D$1024.2D2$
1601.3C2$1601.2C$818.3C2557.2D$818.C2558.D2.D$819.C2557.D2.D$3378.2D$
793.2C977.3C$792.C.C$794.C978.2C1148.D$2179.2D741.D.D$2179.D743.2D$
46.3C190.3C762.3C431.3C739.3D996.D$48.C190.C766.C433.C741.D995.D.D$
47.C192.C764.C433.C1739.D2$1030.2C1748.2D$1030.C.C1747.D.D$1030.C
1750.D$1935.3C619.2D$1935.C622.D$836.3C1097.C618.3D$836.C1718.D$775.
2C60.C$776.2C2865.3C70.C307.2D$775.C2869.C69.2C307.2D$3512.3C70.C58.C
70.C.C$3514.C69.2C259.2D$3513.C70.C.C258.D180.D$2909.3C931.D.D179.D.D
$986.3C1922.C930.D.D181.D$988.C1921.C927.2D3.D183.3D$987.C60.2C1890.
3C895.2D189.D$1047.2C1891.C$1049.C1891.C2$3630.3C90.3C$755.3C92.3C
2870.C$850.C2493.3C70.C81.3C90.3C36.2C91.C$756.2C93.C2494.C69.2C83.C
348.2C$3345.C70.C.C81.C91.2C255.2C$2762.3C1059.2C25.C166.2C$2764.C
1058.C.C191.C.C25.2C$76.2D132.2D2551.C1061.C193.C24.2C$76.2D132.2D
2589.3C1242.C$2801.C$972.3C92.3C1732.C$974.C$973.C93.2C$3147.3C$3149.
C716.2C$1273.2D274.D1598.C716.2C169.3C$1273.D275.3D1670.3C595.2C45.C
168.C$1271.D.D278.D660.3C1006.C596.C.C187.2C26.C$1271.2D278.2D660.C
1009.C597.C186.C.C$2214.C1795.C$750.3C112.3C1291.2C$752.C112.C1292.C.
C$751.C114.C959.D333.C$1824.3D$1823.D$1823.2D910.3C1131.3C$418.2D
2102.3C212.C85.3C1163.3C$417.D2.D2103.C211.C86.C983.2C60.2C120.C$418.
D.D113.D1988.C300.C983.2C180.C$419.D113.D.D421.3C112.3C1502.2C1228.C$
533.D2.D422.C112.C1504.C.C$67.3C15.2C114.2C15.3C313.2D422.C114.C1503.
C$69.C15.C.C112.C.C15.C206.2C$68.C16.C116.C16.C205.C.C3630.2C$425.C
2435.3C118.3C1072.2C$415.3C2445.C118.C145.3C746.3C179.C$417.C121.2C
2321.C120.C146.C746.C198.3C$416.C122.C.C1598.3C986.C666.2C80.C197.C$
539.C753.3C846.C1652.C.C175.2C102.C$529.3C761.C847.C1655.C174.C.C$
531.C716.2C44.C2679.C$530.C718.2C1999.2C$1248.C2001.C.C$3250.C$408.3C
23.2C$434.C.C$409.2C23.C2160.3C$522.3C2070.C$555.2C2039.C1365.2C126.C
$523.2C29.2C3407.2C124.2C$556.C1688.3C1714.C126.C.C$2245.C$2246.C$
1555.3C$54.3C44.2C82.2C44.3C1321.C$101.C.C80.C.C1369.C$55.2C44.C84.C
44.2C1295.3C$116.2C52.2C1358.C$116.C.C50.C.C1357.C$116.C54.C2318.3C$
1818.3C671.C$1583.3C234.C670.C$44.3C194.3C1339.C235.C$1510.3C71.C260.
3C$45.2C194.2C1269.C332.C$1511.C334.C3$1790.3C$1792.C$1791.C71.3C$
1863.C$1864.C220.3C$2087.C1759.2D$545.2D1539.C1756.D3.2D$544.D2.D
1732.3C1559.D.D$430.D113.D.D1733.C1560.D.D189.D$429.D.D113.D1735.C
1559.D191.3D$428.D2.D3408.2D194.D$429.2D3604.D.D$131.2C22.2C381.2C
3496.D$131.C.C20.C.C380.C.C2110.3C$131.C24.C382.C2110.C$547.3C2101.C
1385.2D$424.2C121.C1907.3C1371.2C206.2D$423.C.C122.C1908.C1372.2C$
425.C2030.C1372.C25.2C$433.3C3419.C.C$433.C3421.C$434.C$4024.2C$4023.
C.C$529.2C23.3C3468.C$528.C.C3518.2C$530.C23.2C3493.C.C$440.3C3420.3C
183.C$408.2C3453.C$3C282.3C121.2C29.2C3384.2C36.C$2.C282.C122.C3416.C
.C$.C284.C3540.C3$1470.C2549.2C$1470.2C2547.C.C$1469.C.C2549.C$4064.
2C$1617.3C2444.C.C$1617.C2446.C$1618.C$1905.C1904.2C$1904.2C1905.2C
64.2C$1904.C.C1903.C65.2C$3878.C$1756.3C$1758.C2247.2C$1757.C2249.2C$
4006.C2$1448.3C193.3C$1450.C193.C2427.C$1449.C195.C2425.2C$4071.C.C2$
3899.3C$3899.C$1729.3C193.3C1874.2C96.C$1731.C193.C1875.C.C$1730.C
195.C1876.C2$3991.2C$3990.C.C$3992.C$3791.2C292.2C$3792.2C291.C.C$
3791.C293.C4$3908.2C$3907.2C$3909.C!
-- and a script to build an arbitrary pattern using 90-degree slow glider pairs. He has added some new objects to the recipe list that weren't in the Gemini, notably tub-with-tail eaters. Other objects like the dead spark coil could be added, too -- might want to set up a new forum thread to collect this kind of thing.
simsim314 wrote:Well if I understand you correctly, BLACK and WHITE recipes are totally different, it's just pure chance they both ended in the same lanes. So no, there is no "odd" offset for timing. But maybe I misunderstood your question...
Let me ask it a different way. Your original posting showing a proof-of-concept for an armless Geminoid included one example of two different-colored gliders colliding (4hd separation between the lanes) but also one with same-color gliders (5hd lane separation). Have you found both BLACK and WHITE output gliders by colliding same-color gliders together? Or is there only a VSUC option so far, for 5hd or other same-color glider offsets? I don't know whether to keep the attempts at armless R.U.s that turn out to have same-color outputs.

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 20th, 2014, 4:51 pm

dvgrn wrote:I don't know whether to keep the attempts at armless R.U.s that turn out to have same-color outputs
Oh...I'm checking it. Meanwhile if you see something "good" don't throw it away. My guess is, that in the worst case we end up with five glider instead of four on the "less lucky" color.

Also if we on it: do we have "biased recipes", that have preference for "some color". I mean if the recipe has option to use both white and black with same success it will always use white/black?

User avatar
dvgrn
Moderator
Posts: 10565
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Serizawa - Linear Self Replicator.

Post by dvgrn » April 20th, 2014, 6:09 pm

simsim314 wrote:Also if we on it: do we have "biased recipes", that have preference for "some color". I mean if the recipe has option to use both white and black with same success it will always use white/black?
Nothing like that has been specifically sorted out yet. My general plan for handling optimizations along those lines would be to try gp-compiling as wide a range of recipes as an automated slow-salvo compiler can come up with, and just pick the one that's the shortest as measured in glider pairs.

Here again I'm trying to come up with some standard terminology. A slow-salvo compiler, or "object compiler" maybe, would take a list of constructible objects as input, and produce an optimized slow salvo that builds those objects. Separately from this, a glider-pair compiler (gp-compiler) takes a slow-salvo recipe and turns it into a single stream of glider pairs. I suppose combining those two stages might allow a somewhat better-optimized recipe overall.

Here's a same-color armless design that's only 38sL, and seems like it might have a viable shooting range:

Code: Select all

x = 239, y = 238, rule = LifeHistory
189.A$189.3A$192.A$191.2A7$185.2A$185.2A2$179.2A$180.A$180.A.A$181.2A
2$162.A$162.3A6.2A52.2A$165.A5.2A52.2A$164.2A2$234.2A$165.2A39.2A25.A
.A$165.2A3.2A34.A27.A$170.2A35.3A$209.A$228.2A$198.2A28.2A$152.A7.A
37.2A$140.A11.3A5.3A72.2A$138.3A14.A7.A71.A.A$122.A14.A16.2A6.2A73.A$
122.3A12.2A83.2A13.2A$125.A97.A$124.2A94.3A6.2A$220.A8.2A2$125.2A66.
2A$125.2A17.2A47.A$144.2A48.3A$196.A5$141.2A$141.A$142.3A$144.A$138.
2A$138.A$139.3A$141.A8$128.2A$128.2A12$127.2A$126.A.A$126.A$125.2A9$
137.2A$137.2A6$126.2A$127.A19.2A17.2A$127.A.A17.A18.A$128.2A15.A.A16.
A.A$140.A4.2A17.2A$139.A.A$139.A.A7.2A$128.2A10.A8.2A$127.A.A$127.A$
126.2A38.2A$141.2A16.2A5.2A$141.A17.2A$142.3A$144.A9.A$153.A.A$153.2A
6.2A$124.3A34.A$126.A35.3A$125.A38.A122$2A$.2A$A!
For a diagonally narrow armless R.U. I've been experimenting with designs like this -- back to different-colored colliding gliders:

Code: Select all

x = 265, y = 287, rule = LifeHistory
221.A$219.3A$218.A$218.2A11$194.2A$195.A$195.A.A$196.2A2$225.2A$224.A
2.A$225.2A5$216.2A$216.A.A$218.A$218.2A8$196.2A$195.A.A$195.A$194.2A
7$204.2A$204.2A7.2A$213.A$211.A.A$211.2A3$197.2A$196.A.A$196.A$195.2A
2$251.2A$251.2A3$166.A93.2A$166.3A6.2A82.A.A$169.A5.2A83.A$168.2A43.
2A$213.A.A$215.A38.2A$169.2A44.2A37.2A$169.2A3.2A13.2A$174.2A13.2A70.
2A$261.A.A$263.A$248.2A13.2A$156.A7.A84.A$144.A11.3A5.3A46.2A31.3A6.
2A$142.3A14.A7.A45.2A31.A8.2A$126.A14.A16.2A6.2A19.2A$126.3A12.2A43.A
.A$129.A56.A$128.2A55.2A26.2A$213.2A$199.2A$129.2A67.A.A$129.2A17.2A
48.A$148.2A47.2A8.2A$207.A$208.3A$210.A3$145.2A$145.A$146.3A$148.A$
142.2A$142.A$143.3A$145.A6$151.2A$151.A$132.2A15.A.A$132.2A15.2A12$
131.2A$130.A.A$130.A$129.2A9$141.2A$141.2A6$130.2A$131.A19.2A$131.A.A
17.A$132.2A15.A.A$144.A4.2A$143.A.A28.2A$143.A.A28.A$132.2A10.A27.A.A
$131.A.A38.2A$131.A$130.2A25.2A$145.2A10.2A$145.A$146.3A$148.A25.2A$
167.2A5.2A$167.2A2$162.A$161.A.A$161.2A6.2A$124.3A42.A$126.A43.3A$
125.A46.A122$2A$.2A$A!
This probably has the widest shooting range yet, but it's at the expense of more still lifes and slightly greater diagonal width, so the range may still not be quite wide enough... Also'I haven't really looked at the geometry carefully yet, to be sure that it's going to be useful very often to be able to construct a child R.U. in that orientation.

User avatar
simsim314
Posts: 1823
Joined: February 10th, 2014, 1:27 pm

Re: Serizawa - Linear Self Replicator.

Post by simsim314 » April 20th, 2014, 10:45 pm

OK I've finally made the script working and it has some good results.

Code: Select all

x = 774, y = 776, rule = LifeHistory
771.A$771.A.A$771.2A64$705.A$704.A$704.3A65$638.A$637.A$637.3A79$557.
A.A$557.2A$558.A60$495.A$494.A$494.3A65$428.A$427.A$427.3A78$351.A$
351.2A$350.A.A68$281.A$281.2A$280.A.A69$210.3A$212.A$211.A67$141.A$
141.2A$140.A.A68$71.A$71.2A$70.A.A68$.A$.2A$A.A!
And this:

Code: Select all

x = 783, y = 788, rule = LifeHistory
781.A$780.A$780.3A65$714.A$713.A$713.3A75$637.A$636.A$636.3A79$556.A.
A$556.2A$557.A52$502.A$502.A.A$502.2A73$427.A$426.A$426.3A78$351.A$
351.2A$350.A.A68$281.A$281.2A$280.A.A69$210.3A$212.A$211.A67$141.A$
141.2A$140.A.A68$71.A$71.2A$70.A.A69$3A$2.A$.A!
Now it's not optimized yet for VSUC but it shows that three glider are sufficient for both colors. I'm getting the same result for other d-hd as well (except those that too close to each other, soon will post it all). For now just don't worry about the color, it's not a factor. EDIT Actually I was over optimistic. If we want to stick to 3 collisions, there is only 3hd, 4hd and 5hd are available (3-hd being by far the most potent option). 2hd and 6hd require four collisions , 0hd and 1hd are fully annihilate with after 1-2 collisions.

EDIT: I guess that what we need for any OGC-RU is recipes that can work on both ends. i.e. we want to shoot from left side and right side. Using the same recipe will probably limit one of the sides. So the next step will be to search for two different recipes for each color + side.

Post Reply