Louie, can you give us a chart of exactly how many tests will be done in each k value and how many have been done??????
outlnder
*************
Team Prime Rib
Well, tests done for a specific k value should be Total Tests - Pending Tests, right?
btw. I would change "Remaining tests n > 3000000" (on the project stats page) to "Remaining tests 3M < n < 20M", as the first one would indicate that there are only finite tests left...
Oh, another question:
Are pending tests included in the "Remaining tests n < 3000000" count?
Last edited by Mystwalker; 02-20-2003 at 10:12 AM.
And don't make it a chart, make it text.
In other words, I see the chart on the stats page, but from that I can't figure out exactly how much is left to do. If this is infinite, then just say that.
But if there are finite limits to what we are doing, can we get some type of countdown number for each value??
outlnder
*************
Team Prime Rib
This is an interesting question, not being into mathematics myself.
As far as I understand it, if one of the remaining k's is a Sierpinski number then there will be no prime found for any n of this formula: N = k * 2^n + 1.
So if there is no smarter way to verify that it is infact such a number, this project would last forever.
However, it is expected that all of the k's we are testing for are not Sierpinski numbers. Then the project lasts until we have found the lowest n for each of the k's where N is a prime.
I have not heard anyone made a prediction for in which range on n the primes will be found. It could be soon or not so soon.
There were some "expectations" posted in some threads. Mainly, we were likely lucky to find 5 k/n pairs that fast.
There is no upper limit for testing, as these tests can only proof that it's no Sierpinski number.
So either the other 12 k's get sorted out, too, or we're - well... bust.
That's where this project has its name from. But it should be possible to make a proof that a k is indeed a Sierpinski number - it's just not that easy to find a way...
outlnder, i think the answer to your question is that there are infinite #s left to check for each k... HOWEVER, in reality, there are not an infinite # of tests left... there is just an undeterminable # left.
in other words, we'll know how much work it takes to finish the problem exactly when we finish. it sounds a bit rediculous, but that's the way it goes. there are mathematical predictions that give % chances of finding primes at given levels, but i think they would confuse more people than they would help... because they are very loose and cannot be broken down by k well.
as an example k=65567 should never have had a prime as low as it did. in fact, using nieve projections, that k should have been one of the last three solved, not one of the first three solved.
some people have global projections based on all k value that rely on the fact that the errors in the predictions are averaged out, making them appear slightly better. they are still pretty weak.
anyway, did you notice that i did recently add the work unit counts based on n value? there is a count of tests < 3 mill and another count for tests > 3 mill and < 20 mill right on the overall stats page. there are also the k-graphs you can get by clicking on the # of pending tests that displays how much work is remaining below levels in graph form. those two things capture a good general distribution of where data is.
and just to make it clear, Mystwalker, pending tests are included. the counts are any numbers that aren't flagged 'tested'. pending numbers fall in that category. so when pending tests becomes greater than the < 3 mill count, the natural conclusion would be that numbers greater than 3 mill are begining to be assigned.
-Louie
thx, Louie - good to know.
That means the first n's > 3M will be handed out in approx. 9 days. It would be best to have the lower sieving ranges submitted by then, right?
That means 2x Moo_the_cow. And how's manluver doing?
Last edited by Mystwalker; 02-20-2003 at 05:48 PM.
I should be done in 2 days (with my 655-660 G range).
I'm at 658G now, and I'll submit all of my factors when I'm done.
Oh, 1 more thing:
Louie, could you please post a stats report? I mean, something
like this......
user id , #of factors submitted , #of duplicates
manluver will be done in time.
it would be nice to have the low ones submitted, but it's not critical. i mean, it's not like the sieving will suddenly stop when the first tests greater than n=3 mill go out. the overlap of the sieving ranges and those tests that are assigned won't be very large for the next few months.
-Louie
Moo:
Well, smh said that he accidently computed your range, too, as it was set in his control file when he left for some weeks.
So theoretically he could submit the rest of that range for you...
Louie:
Right, it's not critical, but I think it's best when at least the then already computed parts are submitted. The chance for a factor to divide a n only a tad bigger than 3M is very low, but every one of those saves some guy 50-150 hours (dependant of the computer he uses) of testing...
Smh, submit your factors in the 655-660 range on February 22,
22:22 (UTC)
your sieving range as a double-check.
Sure, but it will be monday.Smh, submit your factors in the 655-660 range on February 22,
22:22 (UTC) . I'll be done by that time, and you can use
your sieving range as a double-check.
It was done with NbeGon btw, so if you are running SoBSieve it will be a nice double check.
Moo, have you submitted your range (655-660G)? I read some discussion in another post.
If so, i'm able to post my results for that range
Oh, I've finished almost all of it (until 659.8 G) but I've only
submitted them until 658.6G. I'll submit the rest of them when
I'm done (which should be done in a couple hours)
Guess i'll be sleeping by then. I'll submit mine tomorrow morning when i'm in the office.
Well, this is the moment you've all been waiting for.
655-660 G is complete, with exactly 200 factors found,
and this means that we have achieved Louie's original
goal of sieving up to 1 T
P.S. Smh, I am using SoBSieve v.1.22, and am using
the latest SoB.dat file.
But i ran this range a month ago, while i was on vacation with a much older SoB.dat file.237 of 237 verified in 86.12 secs.
36 of the results were new results and saved to the database.
I used a version of NbeGon, but dunno which version it was.
If someone wants to check the factors to find out which were new (and duplicate) the file is available on request.
BTW, IIRC version 1.24 of SoBsieve is a bit faster, and the console version adds another 10%
Moo_the_cow, Do you want to try for 2T?
1600 - 1650 Moo_the_cow
Joe O
Posting Permissions
Reply With Quote