mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Data (https://www.mersenneforum.org/forumdisplay.php?f=21)
-   -   where do TF go when they finish? (https://www.mersenneforum.org/showthread.php?t=3779)

markr 2005-03-02 02:01

where do TF go when they finish?
 
1 Attachment(s)
For a while I have been watching the narrowing gap in summary.txt between the leading edges of LL & TF assignments. But it was only recently I noticed something that was right in front of me: the numbers available for LL tests weren't changing, even though TF tests were completing. In fact, in 28M-28.5M they haven't changed since the range was released for LL, except as exponents move from available to out. Example:
[CODE]Status Summary Report 16 Feb 2005 00:00
28200000 28299999 607 18 1543
28300000 28399999 712 16 1258
28400000 28499999 1085 55 710

Status Summary Report 25 Feb 2005 00:00
28200000 28299999 562 20 1543
28300000 28399999 614 24 1258
28400000 28499999 892 69 710[/CODE]The columns with numbers of exponents are TF out, TF factored, and LL available. As these TF complete without finding factors, when do they become available for LL testing? Am I missing something obvious, and this is normal?

I finally got around to looking at the data... summary.txt shows no exponents available for LL tests below 28.4M, so I kept under that. The attachment has all exponents that are in the nofactor file, are less than 28400000, and are not in lucas_v.txt, hrf3.txt, cleared.txt or status.txt. There are 319 below 27.5M that I guess are mostly from manual assignments, but there are 1877 above 27.5M...

Well folks, that's as far as I can go (that is, not far at all) and this post is now way too long, so I'll leave the rest to the collective wisdom out in the forum.

(Details: the GIMPS files I used are from 23/2/2005, the Primenet files have timestamp 01 Mar 2005 22:00, so no server sync in between.)

ET_ 2005-03-02 07:25

I noticed that even when some machine of my team finished their LL work, the total CPU time was not correctly credited. As those machines are far from my location, I just can't send results.txt file to George; I'm wondering if something went wrong during the server outage, pulling exponents out of assignment and now that they are finished can't be credited anymore to the correct ID.

Luigi

garo 2005-03-02 09:54

Typically, exponents that have completed TF are "removed" from Primenet and have to be manually released for LL testing. However, I think it is also possible to mark TF exponents to be released automatically for LL testing. Or maybe there is a daily cron job that does it.

However, since the server has been fudged all bets are off. But yes, LL is catching up to Tf and pretty soon we will have no LL tests left. Then people who request LL tests are likely to be given 10M digit exponents. At least that is what happened he last time we ran out of LL tests due to a reservation script gone awry.

markr 2005-03-02 12:27

My point is just that many completed TF haven't been released for LL in ranges 27.5M and above, since the ranges became available for LL tests - quite a while. In 28M-28.5M at least, I suspect [B]none[/B] of the TF assignments completed since it was opened up for LL testing have become available for LL testing.

Nearly all of the exponents in the list have completed the TF stage, according to the nofactor file. But they have not been LL tested or assigned for an LL test (by PrimeNet anyway), and the leading edge of LL testing has passed them.

The narrowing gap between TF and LL was only how I came to notice this. I shouldn't have opened my post with it since it wasn't my point - you'd think that after all these years I'd be doing better than that!

marc 2005-03-02 12:47

I was under the impression that which numbers were released and how they passed through the various stages of testing was controlled by a human.

I don't *think* a number which is returned from Trial Factoring goes straight into the queue for LL testing. As far as I know George collects all of the trial-factored exponents and then releases them in a batch when LL test is low.

Another possibility is that now server syncs are regular, some exponents get released for LL testing/Double-checking every 15th of the month.

markr 2005-03-09 05:19

Thanks for the replies, folks. The consensus is that it's a manual process - and of course George has had quite a bit of extra stuff to do recently! :rolleyes: But now he has released an extra-large serving of exponents - from summary.txt of 08 Mar 2005 04:00 - about 300 DC, 1600 LL over 10 million digits, and (:drumroll:) 4000 LL under 10 million digits.:bow:


All times are UTC. The time now is 09:10.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.