You're most welcome! You've been doing great work in this thread on the c/10 sub-specialty, mostly all by yourself because nobody else has developed the expertise. It was nice to have a good reason to come and contribute something.Mr. T wrote: ↑November 21st, 2021, 4:37 am@ Goldtiger997, @ jormungant and @ dvgrn
Thank you all very much for reacting to my cry for help! I hope you had fun solving these (at least for me) non-trivial issues.
Now the glorious construct is to shrink. On the strength of your past experience, what method is to prefer for compactification: Giving every edge-shooter its own clock, or using cascading effects with only a few clocks? Most likely both (depending).
Yes, past experience seems to say that you usually end up with a smaller pattern if every edge-shooter has its own gun. This has become gradually more true over the last couple of decades. In the mid-2000s most guns were fairly big Herschel loops, usually with one or more extra gliders that you could get out "for free". Even then, the extra circuits you usually needed to synchronize multiple outputs from one gun, would end up taking up just as much or more space than two separate guns. You can place two guns anywhere you need them, but with two connected streams from one gun there are an annoying lot of constraints.
Occasional exceptions do show up, as you'll find if you scan through the pseudo-guns in the p14 - p77 range in the gun collection.
Recently the average gun has gotten a lot smaller, so that A) it's even cheaper to use two guns instead of one, but on the other hand B) we have much better synchronization circuitry so it's cheaper to reflect and synchronize one stream to match another, but on the gripping hand C) guns are often so optimized that there isn't a spare glider output to take advantage of any more!
The most compact method these days very likely involves building a custom Herschel track so that one Herschel can generate all of the required gliders in more or less the right phases and locations, so that just a few Snarks are needed to bounce them to their respective edge shooters. It would be great to have a search script that could put together candidate circuits for this kind of thing, but it's an absolutely enormous search space and a complex programming challenge -- nobody has dared to tackle it yet.
