Gamedziner wrote:In all honesty, I'd like to find a small stable reflector with ridiculously low repeat time. I'm thinking of using a p20 or less gun plus an eater to search for one using WLS's oscillator search function. To that end, I'd like to know: What is the smallest known glider gun with a (true) period less than or equal to 20?

P20 is the smallest true-period gun known to date. You can tell by checking

chris_c's lists of glider-gun bounding boxes -- if we had a lower-period true gun, we could throw away a huge-bounding-box p14 to p19 pseudogun.

Before you start a WLS search for a p20 glider-stream reflector, read

this estimate about p19 oscillator searches, and then see if you can find a way to constrain your search enough to remove at least 900 zeroes from the time estimate. If your search space is smaller than 20x20, of course that will help proportionally -- e.g., at 14x14, the search time estimate might have "only" about 450 zeroes in it.

Alternatively, to get a sense of how quickly high-period searches can get awful in WLS, start with something simpler: maybe tell WLS about a p20 oscillator --

Code: Select all

```
x = 28, y = 21, rule = B3/S23
4b2o9b2o$3bo2bo2b3o2bo2bo$3b3o2b5o2b3o$b2o3b9o$obo2bo9bo4b2o$obobobo7b
2o3bo2bo$bobobo13bobo2bo2bo$3bo5bo8b2obob5o$3b2o3bobo6bo3bo$3b3o2bo2bo
5bo4bob2o$3b3o5bo5bo3b2o3bo$3b3o2bo2bo5bo4bobobo$3b2o3bobo6bo3bo3bo$3b
o5bo8b2obob2o$bobobo13bobobo$obobobo7b2o3bobo2bo$obo2bo9bo4bo2b2o$b2o
3b9o$3b3o2b5o2b3o$3bo2bo2b3o2bo2bo$4b2o9b2o!
```

-- and then progressively remove more and more known cells from all 20 phases, and see how long it takes WLS to find the oscillator again. With just a few missing cells WLS will be able to work wonders at patching things up, but as more cells become unknown the task gets exponentially more difficult, and at some point you suddenly cross the boundary where you're waiting around for days or weeks to get a result.

Also, without a prime-period true gun I think your proposed method might discover something p2/p4/p5/p10 instead of stable. That would still be a good discovery, but without a lot of clever constraints it seems likely to take billions of years of CPU time to get there.

-- No harm in trying, of course! But your likelihood of success goes way up if you develop a good sense of what size of WLS search can be run to completion.