I have a working code, but it is not good for BOINC. Let me explain.Michael H.W. Weber hat geschrieben: Could you please explain a bit more in detail why you think it might not be worth the effort?
If you come up with a working code, we can run it on Yoyo@home - provided you analyze the resulting data, of course.
Michael.
Not all algorithm can be readily run on BOINC. It needs to be able to run on parallel with minimal data-interdependency, i.e., it can be turned into a form that deals only with small compute range, and combining these ranges will give the whole result.
Unfortunately, the most powerful algorithm up to now does not satisfy this condition. It has to compute a huge hashtable using the whole range, and that defeats most of the attempt to parallelize it. There are algorithms that can be paralellize to give small enough workunits, but the gap of complexity is very large (at least an O(nlogn) factor to be precise). I have done a rough estimate. To check all the parameters up to 7000, using the most powerful algorithm, it can be done with 2 months' computing time on a server with 32 threads; to do the same with less powerful algorithm, it takes at least 7000 times the same computing time on the same machine, therefore at least 7000 such machine on BOINC, and due to the nature of the algorithms, this is only a sub-estimation.
We won't want to do something that can be done on one large server with much more computers. That would be a waste. We can do it in our own computer farm. If we learn something then time is not wasted. But anything put forward to the public needs to be efficient, not to waste.