wildmyron wrote:Unfortunately, I accidentally killed the (9,2)c/11 search that had been running for 3 days [Damn multiple uses of Ctrl-C!!!]. The bounding box used was 13x9, which may be larger than necessary (or too small). I don't think I'll try running it again.
In my experience with the B0 ships, starting with a pretty small search box is a good idea to avoid the solver getting lost.
I suspect what happens sometimes, when the solver just seems to get stuck, is that it gets fixated on a partial solution that is really a dead end, such as a front part that cannot be extended to a full ship, but which the solver cannot definitively rule out because there are too many possible extensions to check. Constraining the search area seems to help the solver get past such dead ends, and maybe find an actual solution instead.
In any case, there seem to be plenty of low-hanging fruit to pick even with pretty narrow searches, and the faster you find those (or find that there's no solution with your constraints) the better. Ideally, you'd like to get either a solution or "Unsatisfiable" in a few minutes at most; if the latter, you can then widen the search box and try again.
What I'm really trying to say here is that I'd guess you probably didn't lose much
useful progress there, if any at all. Just try that search again, this time with a smaller bounding box.
Unfortunately, AFAIK, LLS still handles spaceship searches in a somewhat suboptimal way, where it basically uses a fixed bounding box for the whole search period and then shifts the box by (
x,
y) cells at once. For fast ships with a high period, that creates an awkward "bottleneck" between the last and the first generation.
I modified
my B0 search script to instead gradually slide the bounding box from the original position to the shifted position over the search period, and it seems to really help when looking for fast ships. If you wanted, you could pretty easily edit the script to make it work for non-B0 rules, too.
Actually,
I just did that, and made a bunch of other changes as well while I was at it. The new version now assumes a normal, non-strobing lattice unless explictly told otherwise using the --strobe parameter. You can also tell it to restrict the first generation to a smaller bounding box than other generations, which I've found useful when searching for small ships.