simeks wrote:The database will be two textfiles of this format:
Right -- one (pseudo) still life per line, then, in a format I'm sure I've seen before and that's easy enough to understand anyway. Got it.
I'm sure you're aware of this, but the important part of the verification will be to check that the islands of the results are properly connected through at least one off-cell with more than 3 on-cell neighbours.
Oh, don't count on me being aware of much of anything. My knowledge of cellular automata is -- uneven. Thanks for the tip, though!
The running time for 24 bits is about 4 hours, but I can see a simple way to improve speed by a factor of 2 or 3. The asymptotic speed is around O(2.6^n), the same as reported
by Mark Niemiec.
I'm sure the program can be adapted for other rules, but there are some assumptions about the properties of B3/S23 that will need some consideration.
That's not bad at all - it puts counts up to 28 bits within easy reach.
: actually, thinking about some more now I'm probably not qualified to write a verification tool; I just remembered and looked up the quad pseudo still life
, and I'm not at all sure how you'd detect such patterns as pseudo-objects, short of trying to separate their constituent islands in every possible combination and checking if the results are stable. (After all, while the still lifes making up this particular object only consist of one island each, that needn't be true in the general case.) This would not only take time (naively, O(2^n) for a still life with n islands; granted n would be guaranteed to be quite small for the small objects we're talking about), it might also well exceed my limited programming skills.
So I apologize for having volunteered too eagerly, biting off more than I realized, only to find I couldn't chew it. Someone else will have to do this, someone who actually knows what they're doing.