mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Twin Prime Search (https://www.mersenneforum.org/forumdisplay.php?f=65)
-   -   LLRnet and PRPnet servers for automated LLR (https://www.mersenneforum.org/showthread.php?t=13805)

mdettweiler 2011-01-17 19:18

Now that the rally is concluded and just about everyone's pulled their cores off, it looks like we got as far as n=483335. We need to get it up to n=483999 to dry out the server. Anybody able to spare a computer to help with this? :smile: (My own resources, unfortunately, are booked until the end of January.)

Oddball 2011-01-29 18:20

[QUOTE=mdettweiler;246690]That done, I will load the next range (n=480K-485K, k=100K-200K) into PRPnet as discussed in the "What's next?" thread. [/QUOTE]
For the k=100K-200K part, could you load n=483K-484K instead of n=480K-481K? I'm beginning to suspect that something is wrong with that sieve file (n=483K-485K was originally sieved by pschoefer before being turned over to the distributed effort). It's not the large prime gap from n=482995-483634 that makes me suspicious, it's the fact that there might be three of them in the n=483K-485K range (one from n=482995-483634, one from 483634-484107, and one from n=484406-484996). Even if we do find a prime or two before the PRPnet and LLRnet servers dry out, there will still be an unusually small amount of primes in the n=483K-485K range.

mdettweiler 2011-01-29 21:01

[QUOTE=Oddball;250315]For the k=100K-200K part, could you load n=483K-484K instead of n=480K-481K? I'm beginning to suspect that something is wrong with that sieve file (n=483K-485K was originally sieved by pschoefer before being turned over to the distributed effort). It's not the large prime gap from n=482995-483634 that makes me suspicious, it's the fact that there might be three of them in the n=483K-485K range (one from n=482995-483634, one from 483634-484107, and one from n=484406-484996). Even if we do find a prime or two before the PRPnet and LLRnet servers dry out, there will still be an unusually small amount of primes in the n=483K-485K range.[/QUOTE]
Okay, will do.

Shouldn't be too long now before the PRPnet server dries out and I can upgrade/reload it...should be sometime on Sunday assuming you can stay on until the end. (Otherwise it will be a few more days with just my one quad working on it, plus the core or two from Robert47.)

Lennart 2011-01-31 18:52

[QUOTE=mdettweiler;250349]Okay, will do.

Shouldn't be too long now before the PRPnet server dries out and I can upgrade/reload it...should be sometime on Sunday assuming you can stay on until the end. (Otherwise it will be a few more days with just my one quad working on it, plus the core or two from Robert47.)[/QUOTE]


I'll help you dry it out.:smile:


Lennart

mdettweiler 2011-01-31 18:52

[QUOTE=Lennart;250658]I'll help you dry it out.:smile:


Lennart[/QUOTE]
:bow:

mdettweiler 2011-01-31 21:58

The PRPnet server has now finished drying out. Much thanks to everyone who helped us get it cleaned up! :smile:

Now all we have left of the n=480K-485K, k<100K range is what's currently in the LLRnet server (<24 hours worth of work on a quad).

I'll get to work on upgrading the PRPnet server to version 4.1.4 sometime today. I will, however, need some more sieve files before I can load it with work, and the same goes for LLRnet when it imminently runs out. Grivozavr, can you please post the following files?

n=480K-485K, k=100K-200K (for PRPnet)
n=485K-490K, k<100K (for LLRnet)

Thanks,
Max :smile:

mdettweiler 2011-02-01 06:47

PRPnet server upgrade successful!
 
Hi all,

I am pleased to report that the PRPnet server has been successfully upgraded from version 3.3.6 to 4.1.4 (the latest as of this writing)! :big grin: The upgrade went smoothly and all data is intact.

A couple reminders about what's new and what to keep in mind with the new release:

-4.x has better support for the "Twin" worktype, so we are now using that instead of just "Fixed b, n, c" as before. This means that all -1 primes will have their +1 counterparts tested [b]automatically[/b]. Oddball, you should be able to find the residuals for these supplemental tests in the PRPnet-format daily results files posted at [url]http://www.noprimeleftbehind.net/tps/results/prpnet/[/url] for posting in the primes thread.

Note that I'm not entirely positive yet whether the server will display the candidates on the web page grouped by k or by n (since as of yet I have no sieve file to load into the server for its next range). For the type of work TPS is doing, displaying one web page line for each k is a recipe for disaster whenever anyone tries to load the web page, since there are 50,000 k's in there at a time for only a few n's each. If the server ends up displaying it this way, I'll have to switch it back to "Fixed b, n, c" and put in a feature request for twin-servertype grouping options in a future release. :smile:

-If you are using a 4.1.x client, you can specify a team name under which to be counted in the team stats. You can, of course, leave this blank, in which case your stats will go under team "None". Note that pre-4.1.x clients, which do not support teams, simply show a null entry in the team column (as opposed to even "None") and their results aren't even counted at all in the team stats.

-There is a small bug in this server version that causes the "Team" and "Client" columns to be switched in the User Primes section on the [url=http://noprimeleftbehind.net:12000/all.html]web page[/url]. This is a cosmetic bug only; [b]all the data is still in the right place on the back end[/b]. It will be fixed in the next PRPnet release.

That's pretty much all. Now I just need to wait for gribozavr to send me the new sieve files and we'll be of! :grin:

Max :smile:

gribozavr 2011-02-01 18:35

[url]http://www.sendspace.com/file/owhcei[/url] n_480000-484999__k_100000_200000__01feb2010.txt
[url]http://www.sendspace.com/file/e5f4m2[/url] n_485000-489999__k_1_100000__01feb2010.txt

mdettweiler 2011-02-01 19:35

LLRnet and PRPnet have now been loaded with their new reservations as noted in the LLR Reservation Thread.

To update on what I mentioned yesterday regarding how PRPnet displays Twin-servertype group stats on the web page: I talked with Mark (the PRPnet developer) and he mentioned that while it does group by k by default, only a small code modification is needed to make it group by n. The modification seems to have worked successfully and as such, we will indeed be able to use the Twin servertype (i.e. automatic testing of +1 counterparts to -1 primes) from now on. :smile:

Oddball 2011-02-05 20:48

[QUOTE=mdettweiler;250754]-4.x has better support for the "Twin" worktype, so we are now using that instead of just "Fixed b, n, c" as before. This means that all -1 primes will have their +1 counterparts tested [B]automatically[/B]. Oddball, you should be able to find the residuals for these supplemental tests in the PRPnet-format daily results files posted at [URL]http://www.noprimeleftbehind.net/tps/results/prpnet/[/URL] for posting in the primes thread.
[/QUOTE]
The +1 testing feature is acting rather strangely. There are no residues in the result files, and usernames are missing. Here's what comes out:

[code]
user=mdettweiler
[2011-02-05 07:26:18]
121677*2^483018-1 is prime! Time : 0.0 sec.
user=
[2011-02-05 07:26:18]
121677*2^483018+1 is not prime. Res64: Time : 0.0 sec.
[/code]

The server stats on [URL]http://noprimeleftbehind.net/tps/[/URL] also doesn't display the info properly. Right now, I'm getting:
[code]
Results submitted today by:
mdettweiler : 173
: 1
Total results submitted today: 174
[/code]

mdettweiler 2011-02-05 21:09

Okay, thanks for the heads-up. I need to tweak my LLRnet-format conversion script to handle the different output format used for supplemental tests.


All times are UTC. The time now is 16:51.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.