testitemqlstudop wrote:(By the way, I think I deleted all of my code in rage after plingeling saying unsatisfiable after seven hours. Sorry... I'm thinking of writing another oscillator search program if I have time. In the meantime, if I can tame ofind to run a p19 search...)
Anyone deciding to tackle the omniperiodicity problem in a serious way should probably plan on running searches for something closer to seven months than seven hours. There are a very large number of options to try. If you're lucky, something might show up in seven weeks instead of seven months.
If the search is set up well, I would think the first few thousand options will probably return
unsatisfiable very quickly, so then you'd have to gradually widen the search while keeping each individual sub-search to a known amount of time. When you start hitting searches that don't return a negative result after 100 seconds, maybe put it on a lower-priority queue and eventually come back to it and re-run for 1000 seconds, 10,000 seconds, and so on.
As far as direct p19 oscillator searches go, it will be extraordinarily impressive if you can get a positive result after even seven years of searching. There's a reason why the description of ofind says it "searches for
low-period oscillators."
Apparently
wildmyron's recent p8 oscillator took an overnight run on a reasonably good computer (using JLS). Even if each additional tick in the oscillator period only needed a factor of ten increase in search time (and for an oscillator of any size at all, that's fairly optimistic) then to get a period-19 result you might need
10 hours * 10^11
= about 100 million years of CPU time.
Now, if you come up with an algorithm that no one has ever thought of before, and/or just happen to set up the initial conditions exactly right, it might be only a millionth of that -- but even improving by a factor of a million wouldn't be good enough!
Anyway, if you find a p19, there would still be a period-23 oscillator to discover, which is let's say another ten thousand times harder to search for than the p19 -- and then p38 and p41 beyond that. A working 2c/3 elbow would solve all these problems at once.
