Second that.chris_c wrote:That is very impressive.simeks wrote:Here is a version with just 15 blocks
There are a few blocks added in the construction "danger zone" -- the two NW-most white blocks. @simeks, since your search code found a way to build that key difficult eater in the danger zone, hopefully mere blocks like these can be handled without too much trouble?
Yeah, that's a very nice trick. Once someone gets a decent slow-salvo compiler written for arbitrary Spartan constellations plus all orientations of eater2s, it seems like it might almost be worth re-compiling the 0hd Demonoid, just to see how much smaller it could be made. Definitely worth it if someone else does all the hard work.chris_c wrote:The 0hd-demonoid has two units and we need to destroy the older of the two. Since your reverse glider is on a lane that is transparent to both units it should be possible to fly it straight past the first unit and trigger the destruction in the second one.
Then again, I'm betting on the 19sL replicator unit as the eventual winner of the Smallest Engineered Spaceship competition. It still might turn out to be worth adding a scattering of self-destruct blocks, though, so that the whole thing disappears when you hit it with one or two *WSSes. The square-Orthogonoid design might end up with a much smaller longest diameter, if it can actually be made to work, but it would have a higher population and much longer cycle time, running it in Golly.
-- Come to think of it, if it turns out that the square Orthogonoid is really a viable design, then maybe the analogous trick could be done for a diamond-shaped Demonoid, if it does all its construction and destruction with *WSSes. In which case, that would be my bet for the Smallest Engineered Spaceship award, at least until a cheaper syringe shows up (with "smallest" defined to mean some balance or other between population, bounding box and/or longest diameter).
Ah, that's what's going on. I was impressed with how long it took these patterns to settle, and yet they somehow inevitably boiled away to nothing at the end. Was thinking of a cost function more along the lines of minimizing the remaining population -- but that tends to leave small islands that end up being very expensive to clean up.simeks wrote:The blocks are added one at a time, and after each added block, it is required that the pattern stabilizes into a P2 pattern with no escaping gliders.
...
The patterns with 6 blocks are filtered through a cost function. As the cost I use (width + height) of the bounding box of the pattern once it has settled to a P2 pattern.
The half-perimeter cost function has the huge advantage that all the objects at the edges of the bounding box where the active reaction is, pretty much have to get cleaned up within a few branches of the search tree. So the destruction proceeds in in a fairly orderly way, NW to SE in this case. Very nice!
This would be a great test of my crackpot theory that blinkers have the potential to be more efficient Seeds of Destruction than blocks (because blocks often overpopulate an area and kill off a reaction too soon, where a blinker tends to encourage the reaction to continue.) Beehives might also be an interesting alternate self-destruct seed, since they're not significantly more expensive to build than blocks or blinkers.