Quote:
Originally Posted by jrk
You'd get about half as many relations (I checked). This may necessitate sieving the gaps on the -a side as akruppa suggests.
FYI the relations from my ranges will be posted on Friday afternoon (CST).
|
Okay, how about this:
Using Edh's 80-90M data, I estimate that properly (read: algebraically) sieving the range you provided would get us around 110M rels (and that's the high end of the range, so yield might be better overall). YAFU's minrels estimate for a 30 bit job is 92M rels, and this is on the low end of the 30 bit range, so if we just sieve the remaining gap, that
might be enough to get a half decent matrix. So I'll do just that, and when it's all done I'll attempt a filtering, and do any extra sieving that might be necessary in the ranges I screwed up.
(I'd estimate it to be a 3 day matrix, maybe 4 tops.)
So: I'll (properly) do 38-44M.
PS I said I'd move these back to the 4788 thread, but the tradition seems to be to leave these threads intact. Does anyone care either way?
Edit: With 60-90M and 30-36M rels, or 36M out of the 65M range, there are 46,878,979 unique rels out of 50,087,399 raw (6.4% dup rate). That's with my screwed up rational side regions included. Maybe I really will have to resieve those.