ConwayLife.com

...the state of the art for doing from-zeros searches for fixed board searches like that is the "W cycle avoidance filter" which is a royal pain, performs poorly, and requires a recompilation... Well, I fixed one of those problems anyway. I've just pushed `a6556ac079cf` which adds e.g. `--filters w...

1 TB of RAM and 8 TB of SSDs for swap. Holy crap! If you're taking requests, here are some things I would be interested in (in no particular order): (snip) Some of these might still not be feasible. I'm just throwing ideas out there. I'm sure I'll have more ideas later. You don't have to respond to...

(about c/3 searches in B3/S12) Gutter width 71 hit the first cluster of ships... ...will continue on to odd, even, and asymmetric. Odd width 69 completed with no results in ~32 minutes and maximum of 3.7 GB of memory. Odd width 71 did not make it. Gutter width 71 had hit those first ships at depth ...

(about c/3 searches in B3/S12) In the meantime I have done the recompile to enable the weird filter and have started crawling upwards through the widths. On this highly contested and weak computer where I usually run preliminary searches it has most recently completed width 49 in ~2 minutes and ~13...

Generally for anything LSSS can do it's going to be way, way better than LLSSS at so I can't help but wonder how big/slow its searches run so far have been. I asked for LLSSS specifically because there are known issues with LSSS possibly not being exhaustive, making it useless for disproofs. I have...

Any chance of a search for c/3 orthogonal spaceships in Flock at width 69 gutter? Your program seems like one of the few that's powerful enough to handle that. Filling in the background from the "flock" thread (which is apparently a name for B3/S12) there are known c/3 ships and there is a known wi...

As dreamed of previously and perhaps foreshadowed by some of the EC2 work, I have finally sprung for a new computer with the express purposes of running research computing. It features 1 TB of RAM and 8 TB of SSDs for swap. Yes, eight terabytes, yes just for swap, yes more I think than the combined ...

P.S., Is there a p4 emulator with a spark like 60p7, but with an Ellison side spark? Because of what I’m using it for, a p32 pi/traffic light hassler inspired by the p26 one, I need it to have more clearance than the actual Ellison p4 emulator end. Here’s the mechanism to gauge the required clearan...

After a little more benchmarking my best guess is that "no rcol" is winning where it is due to it benefiting somewhat more from parallelism than the others due to the way the r2l query gets reformulated in its containing loop. All the searches where it was winning were on computers with quite a few ...

This is sort of idle musing into the void, but I was thinking about changing how recentering expansion does its "rcol" index. First, the background: the main structure of the recentering state is "col", a single giant array of pairs of indices into "bcol" (another structure whose contents do not rea...

Here’s a p8 glide-symmetric weak domino sparker partial I made by hand. It is inspired by a sequence that turned up in a collidoscope. x=100,y=40 Rule=B3/S23 4boo$bobobbo$6bo$6bo$obboo$bbo$bbo3bo$6boo$3boobbo$6bo$6bo! If it can be completed, that would be really cool, not only for the sake of its i...

Almost c/7d tubstretcher: x = 50, y = 50, rule = B3/S23 14bo$14bo$13b2o$17b2o$11b3o3bo2$13b3o$14bobo$14b3o2$17b3o$4bo12b2obo$ 4bo11bo2bobo$2bobobo12b3o$3o3b3o$6bobo16bo$7b2o3bo11bobo$3b2o5b2o6b3o b2o2bo$3bo6b2o5bo2bo4bo$10bob2o3bobo2bo$11bobo3b2o2b3o6bo$12b2o6b2o8b o$17bob2o3b3obo2bo$17bo2bo4b2o2bo...

And no luck for common 2c/3 edge dragging a (4, 1)c/8 back. Selected partials: x = 241, y = 21, rule = LifeHistory F29.F29.F29.F29.F29.F29.F.5BA21B.F.5BA21B.F$F29.F29.F29.F29.F.6B21A.F .5B22A.F.5B22A.F.4B23A.F$F29.F29.F.6BA20B.F.5BA21B.F.5BA21B.F.4BA22B. F.4BA22B.F.3BA23B.F$F.6B21A.F.6B21A.F.5B22A.F...

No luck for common 2c/3 edge dragging a 2c/7d back: x = 169, y = 17, rule = LifeHistory F23.F23.F23.F23.F.6B14A2.F.5B15A2.F.5B15A2.F$F2.6BA13B.F2.5BA14B.F2. 5BA14B.F2.4BA15B.F.5BA14B2.F.4BA15B2.F.4BA15B2.F$F2.5B15A.F2.5B15A.F 2.4B16A.F2.4B16A.F.4B16A2.F.4B16A2.F.3B17A2.F$F2.4BA15B.F2.4BA15B.F2. 3BA1...

I've been running that same B2a3i/S12 p3 search on another EC2 box, this one with ~64 GB RAM and ~2 TB of SSDs. It's only a fairly subjective feeling from reading the logs, but it seemed like it was merely "slow" into maybe the ~300-500 GB range and thus that the earlier tests with 8 GB RAM may have...

Gah, if only I had this 4c/8 dot spark (faked here by PASTET EVERY): #C [[ TRACK 0 -1/2 ]] x = 237, y = 174, rule = B3/S23 2bo231bo$b3o5b3o12bo5bo175bo5bo12b3o5b3o$2obo4bo2bo5b3o3b3o3b3o173b3o 3b3o3b3o5bo2bo4bob2o$3o4b3obo4bo2bo3bob2ob2obo173bob2ob2obo3bo2bo4bob 3o4b3o$3o2bo6bo3bo2bo4b3ob3o175b3ob3o...

Some time back I had been wondering if there was a rich world of 3c/4 signals (similar to the rich world of 2c/3 signals), but don't recall much success. IIRC I had found a few 3c/4 tails for still-life fuses, but that was about it. I started looking again but still couldn't find much at merely 3c/4...

Even more 3c/6 heartbreak, 8.5 Y rows of c/2 for blocks edge: x = 427, y = 21, rule = LifeHistory F.43BABABABABABABABABABABABABA.F70.F70.F70.F70.F70.F$F.40B2ABABABABAB ABABABABABABABABA.F.40B2ABABABABABABABABABABABABABA.F.40B2ABABABABABA BABABABABABABABA.F70.F70.F70.F$F.40BA2BABABABABABABABABABABABA...

((24, 12)c/72 "incremental" wickstretcher) Obviously it would be even more ridiculously large, but if you could repeatedly light the wick and stop it burning you could probably swing variable period puffers from this. I had sprayed random cells on the end and was idly watching it burn when I notice...

(about (2, 1)c/6 wickstretchers) Come to think of it, maybe you'd be able to even stretch a wick like this: x = 43, y = 84, rule = B3/S23 o$bo$2bo$3bo$4bo$5bo$6bo$3b3o$2bo$3bo$4bo$5bo$6bo$7bo$8bo$9bo$10bo$11b o$12bo$9b3o$8bo$9bo$10bo$11bo$12bo$13bo$14bo$15bo$16bo$17bo$18bo$15b3o $14bo$15bo$16bo$17b...

NW glider streams are just so very sad. Unfortunately, they are required for the second and third stages because those can't be both unidirectional and simple... I reworked the second stage, now it works at p72 but might be slightly more expensive : x = 228, y = 117, rule = LifeSuper 146.2C$145.C2....

But alas "a little poking in golly" and "simple kickbacks" is about the limit of what I know and I don't personally know how I could pack that SE salvo. I did eventually find some LWSS (W) + glider (NE) -> glider (SE) insertion but I couldn't get it to fit on either side. I finally completed the re...

Are there finite oscillators querying B7c, B8, S5e, S6c, S6e, S6k, S6n? While LLSSS could conceivably provide negative answers here, the first three I looked at all have positive answers, even just among p3 oscillators. S5e I had started with odd symmetry across the initial rows, like: | AABBBuuuCC...

In order to direct further researches, can you summarize the current state of the art of (2,1)c/6 technology, please ? The most important to know for synthesis optimisation would be: - known types of rakes (available directions, edge-shooting ones ...) - garbage deletion for small objects - glider ...

I was dreaming again of trying to make a (2, 1)c/6 wick stretcher for this wick: x = 43, y = 84, rule = B3/S23 o$bo$2bo$3bo$4bo$5bo$6bo$3b3o$2bo$3bo$4bo$5bo$6bo$7bo$8bo$9bo$10bo$11b o$12bo$9b3o$8bo$9bo$10bo$11bo$12bo$13bo$14bo$15bo$16bo$17bo$18bo$15b3o $14bo$15bo$16bo$17bo$18bo$19bo$20bo$21bo$22bo$2...