Prototype / demo for a (weighted) collision search program

For scripts to aid with computation or simulation in cellular automata.
oblique
Posts: 122
Joined: July 16th, 2013, 1:30 pm

Re: Prototype / demo for a (weighted) collision search program

Post by oblique » July 15th, 2014, 1:13 am

Sorry, I did not spend much time near any computer for the last few weeks ...
dvgrn wrote:
dvgrn wrote:With a more expensive way of building the blocks, it's easy to string together a workable 18-glider HWSS seed recipe.
Looks like the lack of a decent HWSS recipe has stalled progress on the 31c/240 ship for the last couple of weeks. Think I should just go ahead and build an HWSS rake with the horribly inefficient 18-glider recipe (which didn't come from the search program, so it will end up having a ridiculously high cost in rephasers)?
With your current set of rakes - and your proven creativity in finding more of those :wink: - those 18 gliders should not be that ridiculously exepnsive ...
dvgrn wrote: -- Or is there another search still in the works that might turn up a more reasonable HWSS seed or two? I suggested cutting the bounding box to maybe 15x15, both to find smaller seeds and to reach out into the COST=15..20 space that hasn't been searched much yet (I don't think).
Not yet. I'll set one up today ... let's see how deep it can dig.
dvgrn wrote: But the Half-Baked Knightship project has been distracting me, so I haven't gotten desperate enough yet to download the latest sscs and try it myself...!
Bitting the user is NYI (not yet implemented), anyway ...
dvgrn wrote: Another question while I'm at it: what exactly does sscs do with patterns that never settle? A new discussion has brought up the question of when the first switch engine makes an appearance in the slow-salvo search tree? Presumably this would be an unweighted search starting from just a blonk (block or blinker), with a bounding box probably at least big enough to hold a pi explosion.
Well. Simply put, it places the reaction-to-analyse in the middle of an empty universe and runs it for a fixed number of generations (the universe is choosen large enough so that even an 1c explosion won't reach the edges).

After those N generations sscs searches for a well defined list of known objects (currently the "natural" space ships) and removes them from the result, backtracking each object to the first generation it appears in.

The reaction is discarded if:

+ an "unwanted" ship appears ("unwanted" is configurable, of corse).
+ the remains after removing any found ships has to big a bounding box.
+ the remains are not (yet) periodic after N generations (or have a higher period then defined as MAX_PERIOD)

Currently, any stuff like puffers or rakes won't fit this search. Even if you'll add the engine of puffer to the list of objects to look for, the ash will not be removed and will mess everything up.

Besides the "objects" my program looks for have to be strictly periodic: still lifes, oscillators, ships.

On the other hand: if we don't care about the remaining result (which would be labled as "pruned" or "unstable", probably) we could define a switch engine as object to look for (bit kludgy, I think ...) - have to think it over ...
dvgrn wrote: Technically it wouldn't matter much if the intermediate reactions emitted gliders, as long as the output gliders didn't get in the way of input gliders -- but that starts to get more complicated to code, unless you just ignore output gliders and check recipes for consistency later. Just the simplest possible search would probably give useful results. I'm betting there's a switch engine within six slow gliders of a blonk -- but now I'm wondering if I'm being too conservative and I should bet on a five-glider recipe instead...!

User avatar
dvgrn
Moderator
Posts: 10612
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Prototype / demo for a (weighted) collision search program

Post by dvgrn » July 15th, 2014, 8:38 am

oblique wrote:Besides the "objects" my program looks for have to be strictly periodic: still lifes, oscillators, ships.

On the other hand: if we don't care about the remaining result (which would be labled as "pruned" or "unstable", probably) we could define a switch engine as object to look for (bit kludgy, I think ...) - have to think it over ...
Hmm. Maybe the right idea is a special-purpose hacked version that only cares about finding the first few infinite-growth patterns, and maybe any particularly impressive methuselahs that showed up along the way -- but really I don't care so much about slow-glider-constructible methuselahs! Starting target has to be just a block or blinker, for Sparse Life compatibility.

It makes sense to use sscs to look for switch-engine seeds -- small periodic objects that you hit with a single glider and get a switch engine. The full search for a minimal infinite-growth Sparse Life configuration is a much harder problem, which Nick Gotts is working on separately.

Seems as if the easy way is just to give each pattern a lot of room, and take the time to run it until it settles. If the population is still increasing after 50K ticks, it's going to be a switch engine. Could even bias the search by sampling the population every 1152 ticks and look for an arithmetic progression (one type of switch engine is period 288 and the other is p384). That probably wouldn't speed things up very much, though, since it would only help with early detection in a tiny percentage of cases.

Paul Chapman found a switch-engine seed just seven slow gliders from a block -- and he only allowed stable intermediate targets, no P2 stuff. So I wouldn't think that a hacked sscs search would have to run very long before it found something, even with the extra overhead from running patterns longer.

Not sure if a switch engine would turn up faster if the intermediate-target bounding box size is made big enough to include pi explosions, or if that will just slow sscs down looking at lots and lots of boring little modifications of pi detritus. I'd kind of guess that a smaller bounding box is better in terms of allowing a wider variety of targets, but something cheaper will probably be hiding somewhere in the larger search space.

oblique
Posts: 122
Joined: July 16th, 2013, 1:30 pm

Re: Prototype / demo for a (weighted) collision search program

Post by oblique » July 16th, 2014, 1:05 am

dvgrn wrote:
oblique wrote:Besides the "objects" my program looks for have to be strictly periodic: still lifes, oscillators, ships.

On the other hand: if we don't care about the remaining result (which would be labled as "pruned" or "unstable", probably) we could define a switch engine as object to look for (bit kludgy, I think ...) - have to think it over ...
Seems as if the easy way is just to give each pattern a lot of room, and take the time to run it until it settles. If the population is still increasing after 50K ticks, it's going to be a switch engine. Could even bias the search by sampling the population every 1152 ticks and look for an arithmetic progression (one type of switch engine is period 288 and the other is p384). That probably wouldn't speed things up very much, though, since it would only help with early detection in a tiny percentage of cases.

[...]

Not sure if a switch engine would turn up faster if the intermediate-target bounding box size is made big enough to include pi explosions, or if that will just slow sscs down looking at lots and lots of boring little modifications of pi detritus. I'd kind of guess that a smaller bounding box is better in terms of allowing a wider variety of targets, but something cheaper will probably be hiding somewhere in the larger search space.
sscs itself is not too well suited for running complicated stuff for thousand of ticks.

In fact, if I took an unmodified sscs and just told it to use MAXGEN = 10000 - it would ask me "OK - where is my 1 TB of main memory?"

I calculate each and every generation first and look for ships in the final one, then backtrack to where the ship started out. (This is way faster, and it does not trip over spurious ships too often). Since I sscs calculates the size of the universe to allow space for lightspeed explosions, the memory consumed is a function (number of generations)^3

And it is in no way optimized for large patterns either.

The reason is that I'm still using the naive one-byte-per-cell-just-check-all-neighbours algorithm I started with. For small patterns that are watch only for a few dozen generations it is good enough to not be the mayor limiting factor.

An easy modification would be to run the current sscs for - say 500 - generations, and then take a deeper look at patterns that are not stabilized after that.

As for this "deeper look":
If we don't care about tracking all those escaping gliders in a "messy" pattern, we would could then switch to a larger universe with only 2 generations and run my naive algorithm over that for another 10k or so.

Or sscs is just used to find candidates, and we'll use bgolly + a nice little script to run those for a few 100k or so.

User avatar
dvgrn
Moderator
Posts: 10612
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Prototype / demo for a (weighted) collision search program

Post by dvgrn » July 24th, 2014, 9:28 pm

oblique wrote:As for this "deeper look":
If we don't care about tracking all those escaping gliders in a "messy" pattern, we would could then switch to a larger universe with only 2 generations and run my naive algorithm over that for another 10k or so.

Or sscs is just used to find candidates, and we'll use bgolly + a nice little script to run those for a few 100k or so.
Yes, if you're using a one-byte-per-cell algorithm, it sounds as if that last option is probably a good one. I'm still curious about how much faster the search might go if it used a more memory-efficient algorithm, though, like the one in Paul Chapman's RLife8.DLL.

Anyway, it seems straightforward to dump to a huge file anything that doesn't settle inside 500 or 1000 ticks. It would be fairly trivial to write a Golly script to go through those candidates until it finds the first infinite-growth pattern. It wouldn't be quick by any means, but it wouldn't take forever, and most likely an unweighted search would eventually turn up a new-record switch engine recipe.

In other news, this morning I took a quick look back through known HWSS recipes for various output directions, and adapted one of them into a four-still-life seed with a fairly impressive 13-cell eastward reach:

Code: Select all

x = 53, y = 145, rule = B3/S23
7b2o5b2o$7b2o4bo2bo$2bo11b2o$obo$b2o2$11b2o$10bo2bo$10bobo$11bo2$4b2o$
4b2o6$25b3o$25bo2bo$25bo$25bo$26bobo5$32b3o$32bo2bo$32bo$32bo3bo$32bo$
33bobo42$39b3o$39bo2bo$39bo$39bo3bo$39bo3bo$39bo$40bobo46$48b3o$48bo2b
o$48bo$48bo3bo$48bo3bo$48bo$49bobo6$39b3o$32b3o4bo2bo$25b3o4bo2bo3bo$
25bo2bo3bo6bo3bo$25bo6bo3bo2bo3bo$25bo6bo6bo$26bobo4bobo4bobo!
Probably it's a bit more expensive than the 18-glider HWSS seed, but it might open up some unlikely options for *WSS salvo build order.

chris_c
Posts: 966
Joined: June 28th, 2014, 7:15 am

Re: Prototype / demo for a (weighted) collision search program

Post by chris_c » July 25th, 2014, 6:30 am

Would something like this work? The lower half of a blockade is used directly in an HWSS seed; the upper half is used to produce a blinker for the next seed. The four non-obvious gliders (gliders 3 to 6) were found with some rather ad-hoc changes to codeholic's monochromatic search program. Throughout the entire construction the only thing that goes further east than the final blinker is a few sparks from the construction of the blockade. I can't imagine that being a show-stopper but is the idea sound in general?

Code: Select all

x = 724, y = 840, rule = B3/S23
o$b2o$2o35$54bo$55bo$53b3o32$40bo$41b2o$40b2o35$94bo$95bo$93b3o27$164b
o$165bo$163b3o3$80bo$81b2o$80b2o35$134bo$135bo$133b3o27$204bo$205bo$
203b3o3$120bo$121b2o$120b2o22$267bo$268bo$266b3o11$174bo$175bo$173b3o
27$244bo$245bo$243b3o27$307bo$308bo$306b3o11$214bo$215bo$213b3o14$356b
o$357bo$355b3o11$284bo$285bo$283b3o27$347bo$348bo$346b3o27$396bo$397bo
$395b3o11$324bo$325bo$323b3o27$387bo$388bo$386b3o$455bo$456bo$454b3o
24$436bo$437bo$435b3o35$525bo$526bo$524b3o3$427bo$428bo$426b3o$495bo$
496bo$494b3o24$476bo$477bo$475b3o35$565bo$566bo$564b3o6$535bo$536bo$
534b3o24$516bo$517bo$515b3o35$605bo$606bo$604b3o6$575bo$576bo$574b3o
43$722b2o$722b2o2$722b2o$722b2o14$645bo$646bo$644b3o6$615bo$616bo$614b
3o3$722b2o$722b2o2$722b2o$722b2o27$722b2o$722b2o2$722b2o$722b2o23$685b
o$686bo$684b3o2$722b2o$722b2o2$722b2o$722b2o27$722b2o$722b2o2$722b2o$
722b2o!

User avatar
dvgrn
Moderator
Posts: 10612
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Prototype / demo for a (weighted) collision search program

Post by dvgrn » July 25th, 2014, 11:54 am

chris_c wrote:Would something like this work?
Absolutely. Going from an 18-glider best recipe to a 7-glider one is quite an improvement! I don't see any problems with the recipe in a 31c/240 context; we'd only be in trouble if we had to build odd-period HWSS streams. (I did have to adjust the one in the link to keep the HWSSes out of the way of the blockade sparks -- it blew up on the first run.)

I think this makes it fairly trivial to complete a 31c/240 caterpillar using codeholic's 90:234:90 front end. (Well, it was fairly trivial before, really, but now the ship will be significantly shorter.) I think the only necessary adjustment will be to move the helix farther away from the block trails, to make room for pairs of fore- and backrakes to generate those initial bi-blocks.

Alternatively, we could create new bi-blocks by pushing the leftover junk to the next upship stream construction lane. Building bi-blocks out of something else turns out not to be terribly expensive:

Code: Select all

#C d3,0 b-3,-3 m-6 ,-3:E9 E15 O-17 E15 O17
#C d3,0 b-7,1 m-10,1:E-9 E-1 O-33 E-1 O1
x = 164, y = 168, rule = B3/S23
obo$b2o$bo49$52bo$50bobo$51b2o64$100bobo$101b2o$101bo32$152bo$150bobo$
151b2o2$157bo$156bobo$156bobo$157bo2$152b2o7b2o$151bo2bo5bo2bo$152b2o
7b2o2$157bo$156bobo$156bobo$157bo!
This might end up being cheaper than colliding forward and backward gliders, just because A) backrakes can be stacked fairly tightly, unlike fore+backrake pairs, and B) it would probably be possible to build support for codeholic's front end without any width-increasing adjustments at all.

On the other hand, we'd need to come up with a lot of custom junk-moving recipes to get from one upship lane to the next, so quite likely the fore+backrake bi-blocks would be cheaper in the end -- both in spaceship length and in total construction time.

So... to make use of codeholic's front end, the next item on the agenda might be to write a script that takes an upship lane number as input, and builds a 90:234:90 side-puffer that puts bi-blocks on that lane. Or it might be worth hunting for more backrakes with different output lanes that work at 90:234:90 first. Some of the same tricks from the 116:110:90 collection will certainly work.

The current 116:110:90 rake collection is attached -- finally got around to adjusting the output gliders so that they're all on the same column, so it's possible to build bi-blocks by connecting any forerake directly to any backrake, as long as the distance between rakes has the right parity. Any number of rephasers can be added between the rakes.

It's still not clear what the most efficient 116:110:90 front end would look like. I'll post a few more ideas when I get a chance, maybe using the new HWSS recipe.
Attachments
31c240-rake-stamp-collection-normalized.mc.gz
Updated 116:110:90 stamp collection with normalized output gliders
(139.52 KiB) Downloaded 419 times

User avatar
codeholic
Moderator
Posts: 1147
Joined: September 13th, 2011, 8:23 am
Location: Hamburg, Germany

Re: Prototype / demo for a (weighted) collision search program

Post by codeholic » July 25th, 2014, 12:41 pm

dvgrn wrote:I think this makes it fairly trivial to complete a 31c/240 caterpillar using codeholic's 90:234:90 front end. (Well, it was fairly trivial before, really, but now the ship will be significantly shorter.) I think the only necessary adjustment will be to move the helix farther away from the block trails, to make room for pairs of fore- and backrakes to generate those initial bi-blocks.
Did I miss something? I don't remember any forerakes for 90:234:90:234:90 :?
Ivan Fomichev

chris_c
Posts: 966
Joined: June 28th, 2014, 7:15 am

Re: Prototype / demo for a (weighted) collision search program

Post by chris_c » July 25th, 2014, 1:28 pm

It's not strictly necessary to go via a bi-block. Here are all of the ways I found to get a "horizontal" hive from the collision of two SE gliders and one NE glider.

Code: Select all

x = 312, y = 19, rule = B3/S23
2bo49bo50bo198bo$obo47bobo48bobo196bobo$b2o48b2o49b2o197b2o3$139bo49bo
$12bo124bobo47bobo$12bo125b2o48b2o$12bo228bo$6b2o49bo49bo131bobo68bo$
5bo2bo47bobo47bobo90bo40b2o67bobo$6bobo47bobo47bobo40bo49bo109bobo$7bo
49bo49bo40bobo48bo110bo$147bo2bo$148b2o45b3o3b3o45b2o$51b2o48b2o51bo
94b2o53b2o$50bo2bo46bo2bo50bo44bo103bo2bo$51b2o48b2o51bo44bo49b2o53b2o
$199bo49b2o!
All of them should maintain the edge shooting capability of the entire construction. In general if the hive is produced within 8 cells of the easternmost cell of any intermediate target then the method should work (if within 9 cells then the method may or may not work).

User avatar
codeholic
Moderator
Posts: 1147
Joined: September 13th, 2011, 8:23 am
Location: Hamburg, Germany

Re: Prototype / demo for a (weighted) collision search program

Post by codeholic » July 25th, 2014, 3:15 pm

dvgrn wrote:In other news, this morning I took a quick look back through known HWSS recipes for various output directions, and adapted one of them into a four-still-life seed with a fairly impressive 13-cell eastward reach:
It seems, it doesn't work with period 240 unfortunately :(

Code: Select all

x = 77, y = 102, rule = B3/S23
2bo$obo$b2o56$67b2o5b2o$67b2o4bo2bo$74b2o4$71b2o$70bo2bo$70bobo$71bo2$
64b2o$64b2o19$67b2o5b2o$67b2o4bo2bo$62bo11b2o$60bobo$61b2o2$71b2o$70bo
2bo$70bobo$71bo2$64b2o$64b2o!
But I wish there was a HWSS recipe with at least some clearance, because that would allow to synthesise an important flotilla, that would reduce the amount of required spaceship salvos in my frontend pattern by 6 (down to 21).

Is there a list of all usable *WSS recipes? If there are good MWSS recipes now, I can try to reproduce my gencols searches (unfortunately I lost almost all data when my hard drive was wiped). That time I excluded MWSS, because there were no known syntheses for them.
Ivan Fomichev

User avatar
dvgrn
Moderator
Posts: 10612
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Prototype / demo for a (weighted) collision search program

Post by dvgrn » July 25th, 2014, 5:57 pm

codeholic wrote:It seems, it doesn't work with period 240 unfortunately :( ...

But I wish there was a HWSS recipe with at least some clearance, because that would allow to synthesise an important flotilla, that would reduce the amount of required spaceship salvos in my frontend pattern by 6 (down to 21).
Drat. Well, I didn't really want to have to slow-construct that ugly four-object constellation, anyway. I'll keep looking. There may be a few more treasures that can be adapted from the old S, E, and W sscs collections.
codeholic wrote:Is there a list of all usable *WSS recipes? If there are good MWSS recipes now, I can try to reproduce my gencols searches (unfortunately I lost almost all data when my hard drive was wiped). That time I excluded MWSS, because there were no known syntheses for them.
I don't have a complete list, but there certainly are quite a lot of good MWSS recipes available now. This one has a large rightward spark, but these two have only small rightward sparks, and there may be a few more that I haven't collected.
codeholic wrote:Did I miss something? I don't remember any forerakes for 90:234:90:234:90.
Now that you mention it, you may not have missed any actual posted patterns. But at least one of the forerake types from 116:110:90 is pretty much guaranteed to work with any block-trail spacing that supports a rephaser:

Code: Select all

#C 3R forerake at 90:234:90:234:90
x = 1005, y = 1184, rule = B3/S23
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o30$
49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o21$
780bo$779b3o$782bo219b3o$780b2o222bo$781bo221bo5$57bo81b2o232b2o88b2o
232b2o$56b3o80b2o232b2o88b2o232b2o$56bob2o$47b2o740b2o$47b2o730b2o8b2o
$779b2o3$48bobo$48b2o$49bo741bo$790b3o$789b2ob2o$788b3o$45b2o742bob2o$
45b2o743b2o$791bo6$49b2o736b2o153b3o$49b2o736b2o155bo$943bo7$373b2o88b
2o237b3o$135b2o236b2o88b2o$135b2ob3o$136b6o553b2o8b2o$57b2o81b3o552b2o
3bo3b3o72b2o$57b2o74bo566bobo2b2o72b2o$132b3o565b2o$131bo2bo$131b3o$
132bo5$143b2o548b2o$143b2o548b2o2$o$obo$2o$882b3o$884bo$49b2o88b2o556b
2o88b2o52bo41bo$49b2o88b2o556b2o88b2o53b2o$841b2o7$373b2o88b2o$373b2o
88b2o3$57b2o72b2o331bo4bo235b2o72b2o$57b2o72b2o330bobo3b3o233b2o72b2o$
458b3o2bobo2bob2o$178bo196b2o80b2obo3b2o3bo2bo$179b2o191bo4bo78bo3bo
10bo$178b2o186b2o4bo5bo67bo5bo3bob2o$365bo2bo3bob2ob2o67bo3b2ob3o3bo
13b2o$104bobo254bo4bob2o15b4o57bo6b2o2b4o11bo2bo$105b2o253bo5bobo6bo8b
2ob2o66bo4bo10b2ob2o364b2o$105bo254b3o2b3o7bo7b2o2bo2bo64b6o9b3o2bo
177bo187b2o$365bo10bo7bo3b3o64bo3b2o7bobo2b2o176b2o187bo$377b2o6bo2bo
2bo63bo2bo8b2obo181b2o$376bo12bobo64b3o9bobo255bo$377b2o10b2o78b2o11bo
243bobo$2b2o479bo242b2o$b2o478b3o$3bo2$49b2o88b2o232b2o88b2o232b2o88b
2o$49b2o88b2o232b2o88b2o232b2o88b2o3$838bo$838bo$838bo2$834b3o3b3o2$
93bo744bo$93bobo742bo$93b2o743bo$57b2o72b2o248b2o72b2o248b2o72b2o$57b
2o72b2o248b2o9bo62b2o248b2o72b2o$393b2o$392b2o356bo$748bobo$749b2o$
780b2o$439bo341b2o$437b2o341bo$438b2o3$62b2o$61b2o$63bo4$49b2o88b2o
232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o3$838bo$838bo$
838bo2$834b3o3b3o2$838bo$838bo$838bo$57b2o322b2o72b2o322b2o$57b2o322b
2o72b2o322b2o3$720b2o$721b2o$720bo4$122b2o$121b2o$123bo6$49b2o88b2o
232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o3$838bo$838bo$
838bo2$834b3o3b3o2$838bo$838bo$838bo$381b2o72b2o$238bo142b2o72b2o$239b
2o419b2o$238b2o421b2o$660bo$301bo$300bo$300b3o290bo$182b2o407b2o$181b
2o409b2o$183bo$542bo$543bo$541b3o5$49b2o88b2o232b2o88b2o232b2o88b2o$
49b2o88b2o232b2o88b2o232b2o88b2o3$838bo$838bo$838bo2$834b3o3b3o2$838bo
$838bo$452b4o382bo$451bo2b2o144b2o$452b2obo145b2o$600bo4$242b2o$241b2o
$243bo10$49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o
88b2o3$838bo$838bo$838bo2$834b3o3b3o2$838bo$838bo$838bo19$57bo81b2o
232b2o88b2o232b2o88b2o$56b3o80b2o232b2o88b2o232b2o88b2o$56bob2o$47b2o$
47b2o789bo$838bo$838bo2$48bobo783b3o3b3o$48b2o$49bo788bo$838bo$838bo2$
45b2o$45b2o4$780bo$780b2o$779bo2bo$49b2o729bobo$49b2o728bo2bo$779bo2bo
$779bobo$780b3o5$373b2o88b2o237b3o$135b2o236b2o88b2o$135b2ob3o$136b6o
553b2o8b2o82b2o$57b2o81b3o552b2o3bo3b3o72b2o8b2o47bo$57b2o74bo566bobo
2b2o72b2o57bo$132b3o565b2o136bo$131bo2bo655b3o$131b3o656b3o41b3o3b3o$
132bo657b4o$792bob2o42bo$791b2ob2o42bo$787b2o2b5o42bo$792b3o$143b2o
548b2o99bo$143b2o548b2o2$o$obo$2o3$49b2o88b2o556b2o88b2o$49b2o88b2o
556b2o88b2o8$373b2o88b2o$373b2o88b2o3$57b2o72b2o331bo4bo235b2o72b2o57b
o$57b2o72b2o330bobo3b3o233b2o72b2o57bo$458b3o2bobo2bob2o366bo$178bo
196b2o80b2obo3b2o3bo2bo$179b2o191bo4bo78bo3bo10bo362b3o3b3o$178b2o186b
2o4bo5bo67bo5bo3bob2o$365bo2bo3bob2ob2o67bo3b2ob3o3bo13b2o363bo$104bob
o254bo4bob2o15b4o57bo6b2o2b4o11bo2bo362bo$105b2o253bo5bobo6bo8b2ob2o
66bo4bo10b2ob2o362bo$105bo254b3o2b3o7bo7b2o2bo2bo64b6o9b3o2bo177bo$
365bo10bo7bo3b3o64bo3b2o7bobo2b2o176b2o$377b2o6bo2bo2bo63bo2bo8b2obo
181b2o$376bo12bobo64b3o9bobo$377b2o10b2o78b2o11bo$2b2o479bo$b2o478b3o$
3bo2$49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o
9$93bo$93bobo$93b2o$57b2o72b2o248b2o72b2o248b2o72b2o57bo$57b2o72b2o
248b2o9bo62b2o248b2o72b2o57bo$393b2o443bo$392b2o356bo$748bobo83b3o3b3o
$749b2o$780b2o56bo$439bo341b2o55bo$437b2o341bo57bo$438b2o3$62b2o$61b2o
$63bo660bobo$724b2o$725bo2$49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b2o
232b2o88b2o232b2o88b2o12$57b2o322b2o72b2o248b2o72b2o57bo$57b2o322b2o
72b2o248b2o72b2o57bo$838bo2$720b2o112b3o3b3o$721b2o$720bo117bo$838bo$
838bo2$122b2o$121b2o$123bo6$49b2o88b2o232b2o88b2o232b2o88b2o$49b2o88b
2o232b2o88b2o232b2o88b2o12$381b2o72b2o381bo$238bo142b2o72b2o381bo$239b
2o419b2o176bo$238b2o421b2o$660bo173b3o3b3o$301bo$300bo537bo$300b3o290b
o244bo$182b2o407b2o245bo$181b2o409b2o$183bo$542bo$543bo$541b3o5$49b2o
88b2o232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o11$452b4o$
451bo2b2o144b2o236bo$452b2obo145b2o235bo$600bo237bo2$834b3o3b3o2$242b
2o594bo$241b2o595bo$243bo594bo10$49b2o88b2o232b2o88b2o232b2o88b2o$49b
2o88b2o232b2o88b2o232b2o88b2o12$838bo$838bo$838bo2$834b3o3b3o2$838bo$
838bo$838bo10$57bo81b2o232b2o88b2o232b2o88b2o$56b3o80b2o232b2o88b2o
232b2o88b2o$56bob2o$47b2o$47b2o4$48bobo$48b2o$49bo3$838bo$45b2o791bo$
45b2o791bo2$834b3o3b3o$783b3o$782bo2bo52bo$782b2o2bo51bo$785b2o51bo$
49b2o$49b2o731bo$780b2o$781b2o6$373b2o88b2o237b3o$135b2o236b2o88b2o$
135b2ob3o$136b6o553b2o8b2o$57b2o81b3o552b2o3bo3b3o72b2o$57b2o74bo566bo
bo2b2o72b2o$132b3o565b2o$131bo2bo$131b3o$132bo4$838bo$143b2o548b2o143b
o$143b2o548b2o143bo2$o833b3o3b3o$obo$2o836bo$838bo$838bo$49b2o88b2o
556b2o88b2o$49b2o88b2o556b2o88b2o8$373b2o88b2o$373b2o88b2o3$57b2o72b2o
331bo4bo235b2o72b2o$57b2o72b2o330bobo3b3o233b2o72b2o$458b3o2bobo2bob2o
$178bo196b2o80b2obo3b2o3bo2bo$179b2o191bo4bo78bo3bo10bo$178b2o186b2o4b
o5bo67bo5bo3bob2o$365bo2bo3bob2ob2o67bo3b2ob3o3bo13b2o$104bobo254bo4bo
b2o15b4o57bo6b2o2b4o11bo2bo$105b2o253bo5bobo6bo8b2ob2o66bo4bo10b2ob2o$
105bo254b3o2b3o7bo7b2o2bo2bo64b6o9b3o2bo177bo$365bo10bo7bo3b3o64bo3b2o
7bobo2b2o176b2o$377b2o6bo2bo2bo63bo2bo8b2obo181b2o$376bo12bobo64b3o9bo
bo$377b2o10b2o78b2o11bo$2b2o479bo$b2o478b3o$3bo2$49b2o88b2o232b2o88b2o
232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o9$93bo$93bobo$93b2o$57b2o
72b2o248b2o72b2o248b2o72b2o$57b2o72b2o248b2o9bo62b2o248b2o72b2o$393b2o
$392b2o356bo$748bobo$749b2o$780b2o$439bo341b2o$437b2o341bo$438b2o3$62b
2o$61b2o$63bo3$722bo$49b2o88b2o232b2o88b2o232b2o21b2o65b2o$49b2o88b2o
232b2o88b2o232b2o22b2o64b2o12$57b2o322b2o72b2o248b2o72b2o$57b2o322b2o
72b2o248b2o72b2o3$720b2o$721b2o$720bo4$122b2o$121b2o$123bo6$49b2o88b2o
232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o12$381b2o72b2o$
238bo142b2o72b2o$239b2o419b2o$238b2o421b2o$660bo$301bo$300bo$300b3o
290bo$182b2o407b2o$181b2o409b2o$183bo$542bo$543bo$541b3o5$49b2o88b2o
232b2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o11$452b4o$451bo
2b2o144b2o$452b2obo145b2o$600bo4$242b2o$241b2o$243bo10$49b2o88b2o232b
2o88b2o232b2o88b2o$49b2o88b2o232b2o88b2o232b2o88b2o!
It would be sheer luck, but the odds aren't too bad that there's a 2R version. The trick would be to change the right-hand glider timings of the first following rephaser to create a small explosion near the traffic light that happens to destroy it cleanly.

That's just a matter of trying each of the 60-odd possible collisions (30+ same color and 30+ opposite color) in each of the two possible phases relative to the traffic light. Anyone interested in a small research project?

EDIT: 90:234:90 has enough room to allow significant +/-31N adjustments to the lane of the two-glider suppression/collision/construction, if that turns out to be any use. Here's an extreme example:

Code: Select all

#C sample 90:234:90 bi-block and interchange side-puffers
x = 885, y = 1849, rule = B3/S23
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$
34b2o88b2o232b2o88b2o232b2o$34b2o88b2o232b2o88b2o232b2o$770bo$769b2o$
767b2obo$774bo$766bo2bo4b2o$767b2obobo2b2o$766bobo2b2o$766bo2bo$766bo$
767bobo$768bo2$776b2o$776b2o7$772b2o$772b2o8$34b2o88b2o232b2o88b2o$34b
2o88b2o232b2o88b2o3$679b3o5bo76b2o$687b2o75b2o44bo$679b3o4b2o3bo119bo$
679b3o9bo117b3o$691bo6$678b2o$678b2o7$682b2o88b2o$682b2o88b2o7$729bo$
34b2o88b2o232b2o88b2o277b2o$34b2o88b2o232b2o88b2o199bo78b2o$648bo$648b
3o$690b2o72b2o$690b2o72b2o17$682b2o88b2o$682b2o88b2o3$722bo$723b2o$
722b2o3$34b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o3$690b2o72b2o$690b
2o72b2o17$682b2o88b2o$682b2o88b2o8$34b2o88b2o232b2o88b2o416b2o$34b2o
88b2o232b2o88b2o416b2o2$866b2o$764b2o100b2o$764b2o17$682b2o88b2o$682b
2o88b2o5$809b3o$811bo$589bo220bo$34b2o88b2o232b2o88b2o138bo277b2o$34b
2o88b2o232b2o88b2o138b3o275b2o2$866b2o$866b2o18$682b2o88b2o$682b2o88b
2o3$749b3o$751bo$750bo3$124b2o232b2o88b2o416b2o$124b2o232b2o88b2o416b
2o$32bo$32bo5bo827b2o$31b3o5bo826b2o$37bobo$31b3o4b2o$32bo9b3o7$30b2o$
30b2o7$34b2o646b2o88b2o$34b2o646b2o88b2o$689b3o$691bo$690bo5$358b2o88b
2o416b2o$2bo120bob2o231b2o88b2o416b2o$2o115b2o4bo2bo$b2o115bo4bo742b2o
$42b2o75b3obo742b2o$42b2o75bo2b2ob3o$119bo3bob3o$121b2o4$106b5o$107b3o
$108bo20b2o$128bo2bo6bo$128bob2o4b2o$129bo3b5o$133bo2bo$130bob4o$131bo
$130bo2$34b2o88b2o22bo480b3o50b2o88b2o$34b2o88b2o23bo481bo50b2o88b2o$
147b3o480bo2$75bo$73bobo$74b2o453bo$528bo$528b3o$358b2o88b2o$358b2o88b
2o3$42b2o72b2o$42b2o72b2o7$444b3o$443bo2bo$442bo4bo$444bo$95bo$93b2o
350b2o$94b2o348bobo$366bo77bobo$363b6o200b3o$363b2o2b2o81bo120bo$34b2o
88b2o322bo121bo111b2o88b2o$34b2o88b2o238bo83b2o232b2o88b2o$364bo2b2o$
363bo3b2o$364bo74bo$364bobo71bo$365bo72b3o6$2b3o351b2o$2bo39b2o72b2o
237bo10b2o72b2o$3bo38b2o72b2o237b3o8b2o72b2o$355bo$354bo2bo2$355bobo9$
509b3o$511bo$510bo2$34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o232b2o
88b2o232b2o88b2o9$62b3o$62bo$63bo$42b2o322b2o72b2o$42b2o322b2o72b2o15$
208bo$209bo$34b2o88b2o81b3o148b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b
2o232b2o88b2o2$301bobo$301b2o$302bo3$122b3o$122bo$123bo3$366b2o72b2o$
366b2o72b2o13$422bo$423bo$421b3o2$34b2o88b2o232b2o88b2o232b2o88b2o$34b
2o88b2o232b2o88b2o232b2o88b2o$379bo$378bo$378b3o2$182b3o$182bo$183bo5$
366b2o72b2o$366b2o72b2o17$34b2o88b2o232b2o88b2o232b2o88b2o$34b2o88b2o
232b2o88b2o232b2o88b2o3$242b3o$242bo$243bo25$34b2o88b2o232b2o88b2o232b
2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$34b2o88b2o232b2o88b2o232b2o
88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$34b2o88b2o232b2o88b2o232b2o
88b2o$34b2o88b2o232b2o88b2o232b2o88b2o30$34b2o88b2o232b2o88b2o232b2o$
34b2o88b2o232b2o88b2o232b2o$770bo$769b2o$767b2obo$774bo$766bo2bo4b2o$
767b2obobo2b2o$766bobo2b2o$766bo2bo$766bo$767bobo$768bo2$776b2o$776b2o
7$772b2o$772b2o8$34b2o88b2o232b2o88b2o232b2o$34b2o88b2o232b2o88b2o232b
2o3$764b2o$764b2o44bo$811bo$809b3o15$772b2o$772b2o7$729bo$34b2o88b2o
232b2o88b2o277b2o$34b2o88b2o232b2o88b2o278b2o3$679b3o5bo76b2o$687b2o
75b2o$679b3o4b2o3bo$679b3o9bo$691bo6$678b2o$678b2o7$682b2o88b2o$682b2o
88b2o7$729bo$34b2o88b2o232b2o88b2o277b2o$34b2o88b2o232b2o88b2o199bo78b
2o$648bo$648b3o$690b2o72b2o$690b2o72b2o17$682b2o88b2o$682b2o88b2o3$
722bo$723b2o$722b2o3$34b2o88b2o232b2o88b2o$34b2o88b2o232b2o88b2o2$870b
o$690b2o72b2o105bo$690b2o72b2o103b3o17$682b2o88b2o$682b2o88b2o7$869b3o
$34b2o88b2o232b2o88b2o421bo$34b2o88b2o232b2o88b2o420bo3$764b2o$764b2o
6$880b3o2$878bo5bo$878bo5bo$878bo5bo5$878bo5bo$878bo5bo$682b2o88b2o
104bo5bo$682b2o88b2o$880b3o4$809b3o$811bo$589bo220bo$34b2o88b2o232b2o
88b2o138bo$34b2o88b2o232b2o88b2o138b3o10$880b3o2$878bo5bo$878bo5bo$
878bo5bo5$878bo5bo$878bo5bo$682b2o88b2o104bo5bo$682b2o88b2o$880b3o2$
749b3o$751bo$750bo3$124b2o232b2o88b2o$124b2o232b2o88b2o$32bo$32bo5bo$
31b3o5bo$37bobo$31b3o4b2o$32bo9b3o7$30b2o$30b2o7$34b2o646b2o88b2o$34b
2o646b2o88b2o$689b3o$691bo$690bo5$358b2o88b2o$2bo120bob2o231b2o88b2o$
2o115b2o4bo2bo$b2o115bo4bo$42b2o75b3obo$42b2o75bo2b2ob3o$119bo3bob3o$
121b2o4$106b5o$107b3o$108bo20b2o$128bo2bo6bo$128bob2o4b2o$129bo3b5o$
133bo2bo$130bob4o$131bo$130bo2$34b2o88b2o22bo480b3o50b2o88b2o$34b2o88b
2o23bo481bo50b2o88b2o$147b3o480bo2$75bo$73bobo$74b2o453bo$528bo$528b3o
$358b2o88b2o$358b2o88b2o3$42b2o72b2o$42b2o72b2o7$444b3o$443bo2bo$442bo
4bo$444bo$95bo$93b2o350b2o$94b2o348bobo$366bo77bobo$363b6o200b3o$363b
2o2b2o81bo120bo$34b2o88b2o322bo121bo111b2o88b2o$34b2o88b2o238bo83b2o
232b2o88b2o$364bo2b2o$363bo3b2o$364bo74bo$364bobo71bo$365bo72b3o6$2b3o
351b2o$2bo39b2o72b2o237bo10b2o72b2o$3bo38b2o72b2o237b3o8b2o72b2o$355bo
$354bo2bo2$355bobo9$509b3o$511bo$510bo2$34b2o88b2o232b2o88b2o232b2o88b
2o$34b2o88b2o232b2o88b2o232b2o88b2o9$62b3o$62bo$63bo$42b2o322b2o72b2o$
42b2o322b2o72b2o15$208bo$209bo$34b2o88b2o81b3o148b2o88b2o232b2o88b2o$
34b2o88b2o232b2o88b2o232b2o88b2o2$301bobo$301b2o$302bo3$122b3o$122bo$
123bo3$366b2o72b2o$366b2o72b2o13$422bo$423bo$421b3o2$34b2o88b2o232b2o
88b2o232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o$379bo$378bo$378b3o2$
182b3o$182bo$183bo5$366b2o72b2o$366b2o72b2o17$34b2o88b2o232b2o88b2o
232b2o88b2o$34b2o88b2o232b2o88b2o232b2o88b2o3$242b3o$242bo$243bo!

User avatar
dvgrn
Moderator
Posts: 10612
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Prototype / demo for a (weighted) collision search program

Post by dvgrn » August 22nd, 2014, 4:21 pm

With further progress recently on a 116:110:90 front end, it looks to me as if a 31c/240 spaceship is within easy reach now.

The only problem is that it looks as if a forerake slow salvo will allow for a very much shorter ship. Otherwise we have to wait around a ridiculous length of time (and therefore length of spaceship) for backrake gliders to reach the faraway HWSS streams.

But forerake constructions are pretty much unknown territory. I had one more look through the latest sample_e and sample_s patterns, but didn't turn up any workable HWSS seeds that could be constructed with forerakes. Am I missing anything?

Unless something better comes along in the next couple of weeks, I'll probably just cobble together a recipe for two seeds along these lines --

Code: Select all

x = 359, y = 683, rule = B3/S23
333bo$332bobo$333b2o4$337b2o$337b2o2b2o$341b2o12$343b3o11$333bo$332bob
o$333b2o4$337b2o$337b2o2b2o$341b2o12$343b3o11$333bo$332bobo$333b2o4$
337b2o$337b2o2b2o$341b2o12$343b3o11$333bo$332bobo$333b2o4$337b2o$337b
2o2b2o$341b2o12$343b3o11$333bo$332bobo$333b2o4$337b2o$337b2o2b2o$341b
2o12$343b3o11$333bo$332bobo$333b2o4$337b2o$337b2o2b2o$341b2o12$343b3o
13$341bo$300b2o$301b2o53b3o$300bo54bo2bo$358bo$341bo12bo3bo$354bo3bo$
358bo$355bobo2$341bo2$350bo$349b3o$349bob2o$341bo8b3o$350b3o$350b3o$
350b2o2$341bo5$341bo5$240b2o99bo$241b2o$240bo3$341bo5$341bo5$341bo5$
341bo5$341bo4$180b2o$181b2o158bo$180bo4$341bo5$341bo5$341bo5$341bo5$
341bo3$120b2o$121b2o$120bo220bo$356b3o$355bo2bo$358bo$354bo3bo$341bo
12bo3bo$358bo$355bobo3$341bo5$341bo5$341bo5$341bo2$60b2o$61b2o$60bo$
341bo5$341bo5$341bo5$341bo5$341bo5$341bo$2o$b2o$o2$341bo5$341bo4$340bo
$339bobo$340b2o4$344b2o$344b2o2b2o$348b2o11$351bo$351bo$351bo4b3o$355b
o2bo$358bo$354bo3bo$354bo3bo$358bo$355bobo4$340bo$339bobo$340b2o4$344b
2o$344b2o2b2o$348b2o11$351bo$351bo$351bo10$340bo$339bobo$340b2o4$344b
2o$344b2o2b2o$348b2o11$351bo$351bo$351bo10$340bo$339bobo$340b2o4$344b
2o$344b2o2b2o$348b2o9$356b3o$355bo2bo$351bo6bo$351bo2bo3bo$351bo2bo3bo
$358bo$355bobo8$340bo$339bobo$340b2o4$344b2o$344b2o2b2o$348b2o11$351bo
$351bo$351bo10$340bo$339bobo$340b2o4$344b2o$344b2o2b2o$348b2o11$351bo$
351bo$351bo12$308b2o$307bobo$309bo9$356b3o$355bo2bo$358bo$354bo3bo$
354bo3bo$358bo$355bobo12$248b2o$247bobo$249bo27$188b2o$187bobo$189bo
27$128b2o$127bobo$129bo27$68b2o$67bobo$69bo27$8b2o$7bobo$9bo!
-- maybe with a different glider turner, if something cheaper shows up. The dotted lines show where the single leftover block has to be slow-converted into the two blocks, boat, and blinker constellation.

Can anyone come up with a better mechanism -- or a complete forward slow salvo for this design? Unfortunately I'm booked pretty solid for the rest of this month.

User avatar
dvgrn
Moderator
Posts: 10612
Joined: May 17th, 2009, 11:00 pm
Location: Madison, WI
Contact:

Re: Prototype / demo for a (weighted) collision search program

Post by dvgrn » September 4th, 2014, 10:55 am

dvgrn wrote:With further progress recently on a 116:110:90 front end, it looks to me as if a 31c/240 spaceship is within easy reach now.
We have liftoff: the Shield Bug has launched.

It doesn't look anything like what I expected it to at the beginning of the project, or during the big sscs backrake searches. When a new, "simpler" front end showed up, it turned out that the support *WSSes had to be very very far away from the block trails. That effectively meant that just about any design that constructed the upship streams with backrake slow salvos would end up being maybe three times longer than any design using forerakes.

The key HWSS seed had turned up in sscs searches, but had previously been found by codeholic. So unfortunately sscs hasn't really had its 15 nanoseconds of fame... yet! The forward-glider salvo for the HWSS seed was compiled by hand, so very likely there's something within reach of sscs that's about half as long -- six or seven forerakes, maybe, instead of 11? It's clearly possible to build a much smaller 31c/240 ship, so it would be nice to have really minimal forerake *WSS-seed recipes for it.

This is especially true now that more new forerakes are showing up. I'll post an updated collection here when I get a chance.

User avatar
biggiemac
Posts: 515
Joined: September 17th, 2014, 12:21 am
Location: California, USA

Re: Prototype / demo for a (weighted) collision search program

Post by biggiemac » September 20th, 2016, 7:18 pm

Well, I've now read through the entirety of this thread and of the 31c/240 working notes thread seeing if any of it is directly applicable to the (27,1)c/72 effort. The major differences: (27,1)c/72 is oblique, faster-than-gliders, and unable to operate at x1.

For the left side of the ship, where slow salvos are applicable, obliquity has a couple ramifications. While slow salvos get a bit more clearance for free, any debris is problematic. The "lane" of the construction product is not a vertical line, and so intermediate targets may be wherever they like. After releasing the *WSS, the debris is immediately a hazard to following ships in the stream. For that reason, any *WSS construction done by the singleton constructor (which consists of monochromatic p4 NW forerakes) should produce no debris, or if any it must be far enough away from the released *WSS to be cleaned up by the next rake. If another construction is to follow, the ideal object for the pattern to leave behind in addition to the *WSS is a NE glider.

I understand that glider outputs were removed from the scss search by default, since they are not compatible with true slow salvo. However, what I would really like to find is a clean slow-cheap converter for a NW glider to a NE glider and edgy N LWSS. Something like the following, just with an associated recipe:

Code: Select all

x = 31, y = 17, rule = B3/S23
15b2o$15b2o$3b2o$3b2o3$7b2o$7b2o11b2o$19bo2bo$19bobo6b2o$20bo7bobo$28b
o3$b2o$o2bo$b2o!
"Cost" is really only measurable in terms of number of rakes - for the singleton stream all lanes can be reached within 26 vertical cells of the most convenient lane. For now I am looking through the last samples_s pattern file in this thread to see if there are easily modified options (like an edge shooter with an internally deleted glider which can be freed by deleting one object). I'm not sure if the monochromatic restriction invalidates the existing cost function.

This also neglects to mention any of the other tricks applicable - like how chris_c made the HWSS recipe for the 31c/240 projects use the top of one site as the bottom of the next site. Factoring in a (27,1) skewed-cylinder wrap is probably a lot of trouble to go to. Another trick is the parallel construction of a NE-glider-only pattern, necessary since this is unable to operate at x1. The whole issue of parallel constructions branching, reconverging, and rebranching is enough to make my head hurt.

The initial targets can vary, depending on when the parallel construction branches. For now just starting from a block or HF (either color) should give applicable results.

There's so much overlap between different caterpillar projects, and yet so little at the same time..!
Physics: sophistication from simplicity.

Post Reply