mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Data

Reply
 
Thread Tools
Old 2003-09-12, 17:33   #1
GP2
 
GP2's Avatar
 
Sep 2003

13·199 Posts
Default Runaway trains

The file LUCAS_V.TXT contains verified-good results.

Usually, each exponent appears there twice: once for the original LL test, once for the double-check. But a few exponents appear multiple times. Note verified-bad results are stored in a separate file (BAD), so every time an exponent appears more than twice in LUCAS_V.TXT it represents a redundant double-check of an already verified exponent.

Which exponents appear the largest number of times? Here are the champions:

305 6522911 (TempleU-DI,AL19)
244 8893783 (TempleU-DI,FL-SLE)
73 5721461 (TempleU-DI,AL22)
54 4983379 (kgymnas,CBA6E3613)
44 9368201 (TempleU-DI,all)
36 13283969 (Northwestern,C1942B4AC)
28 8883257 (TempleU-DI,C7B17693A)
26 12178097 (telos,CC8FC13C8)
24 11636461 (Northwestern,C866EF1BC)
20 2397103 (S07796,)
20 18491911 (WillametteMath,cln-411-*) (multiple machines)
15 9633553 (vanz,*) (multiple machines)

In most of these cases, a single computer returned these results over and over again.

Think of all the wasted cycles.

Sometimes a repeated test by the same computer is the result of restoring a backup, which results in an older version of worktodo.ini. This is unlikely to happen more than once however for any given exponent. So the above cases are probably the result of making files like worktodo.ini non-writeable, or perhaps a filesystem out of disk space preventing writing new versions of files, which amounts to the same thing.


Fortunately, most of the above runaway trains seem to have been stopped by now (they are no longer adding new lines to LUCAS_V.TXT). There seem to be very few currently active runaway trains.

These have added new redundant results between July 30 and Sept 9:

7860253,Northwestern,C5BD0A501,WW1 (7 times total)
14417057,S104512,C2D476D4E,WX5 (14 times total)
18445951,telos,gx150,WX5 (4 times total)

They're running old versions of the software (current version 23 would be WZ1 or WZ2) and may be machines the users no longer have access to. Still... anyone who knows the e-mail addresses of these folks want to give them a shout?

Maybe Prime95 and mprime can be made more robust to prevent this, but it's not a major issue. Percentage-wise, it's a very, very tiny fraction of the 529877 total lines in LUCAS_V.TXT . And the worst of the problem seems to be in the past.
GP2 is offline   Reply With Quote
Old 2003-09-12, 21:34   #2
garo
 
garo's Avatar
 
Aug 2002
Termonfeckin, IE

53138 Posts
Default

As you can see most of those returning these multiple results are big teams, University teams. I belive it is not the same computer returning the same result but the case of several computers being imaged with a non-empty worktodo.ini. In this case all the computers that are images will have the same values in the .ini files and would appear to be the same computer.
garo is offline   Reply With Quote
Old 2003-09-13, 00:41   #3
GP2
 
GP2's Avatar
 
Sep 2003

13×199 Posts
Default

Hmmm, I hadn't thought of that (new computers getting imaged).

Another thing is that some of these large teams are probably still using older versions of the software. And if they copy images to new machines as you suggest, they're simply copying the older version.

Maybe we should try to identify large teams using older versions and get the powers-that-be to send them an email inviting them to upgrade to version 23? Large cash bribes would ensure that we forget to notify curtisc.


Maybe the server code could be edited to raise a flag every time a P4 returns a result from a version earlier than 23, and target those folks
for e-mail?

If all the existing computers become 30% more efficient, it's the same as if we recruited 30% more users.
GP2 is offline   Reply With Quote
Old 2003-09-13, 01:47   #4
PrimeFun
 
PrimeFun's Avatar
 
Jul 2003

2·3·5 Posts
Default

Keep in mind large institutions may not update their system images as often as you would think. After all, system images are created to reduce the time it takes get a system up and running with a stable known quantity or used as an emergency recovery device. Some student labs are restored back to a previous version on a regular schedule.

Another factor to keep in mind is the benefit of Prime95 staying low and under the radar. I know there has been discussions on the mailing list before where individuals have talked about getting Prime95 included in a system image of their institutions by begging and pleading the IT guys. If those same IT guys aren't still in power or just don't really care anymore, bringing it up again could turn a yes into a no.

As for applying updates system wide, while it is a pretty easy thing to do the individuals who care enough about Prime95 to go through the trouble are probably the same type of people that keep up to date on the happenings of Prime95 to already be aware.

I'm not saying it's a bad idea; just one that is slightly more complicated than it appears on the surface.
PrimeFun is offline   Reply With Quote
Old 2003-09-13, 02:23   #5
outlnder
 
outlnder's Avatar
 
Aug 2002

13E16 Posts
Default

I would hope that Tyemple Univ. is not getting credit for all those redundant results??
outlnder is offline   Reply With Quote
Old 2003-09-13, 08:21   #6
garo
 
garo's Avatar
 
Aug 2002
Termonfeckin, IE

1010110010112 Posts
Default

Not on primenet but yes on George's score list.
garo is offline   Reply With Quote
Old 2003-09-14, 02:37   #7
nomadicus
 
nomadicus's Avatar
 
Jan 2003
North Carolina

2·3·41 Posts
Default

Keep in mind if files aren't restored correctly, even a read-only on the results.txt file can cause this to happen.
nomadicus is offline   Reply With Quote
Old 2003-09-14, 18:53   #8
markhl
 
Apr 2003
California

22×23 Posts
Default

Quote:
Originally posted by nomadicus
Keep in mind if files aren't restored correctly, even a read-only on the results.txt file can cause this to happen.
This can happen if e.g. you copy files from a CD-R that you used to backup or move files.

Windows (all versions?) sees files on a closed CD-R as read-only, and preserves that flag when they are copied to hard disk. So right-click the folder, choose Properties, & clear the read-only flag.
markhl is offline   Reply With Quote
Old 2003-10-26, 04:55   #9
GP2
 
GP2's Avatar
 
Sep 2003

13·199 Posts
Default

Is Scott Kurowski still around?

He has 20 entries for exponent 20494237 in LUCAS_V.TXT.

The odd thing is that those 20 entries appeared all at once in the Oct 9 version of the data files. Prior to that there were no entries at all.

So is this a runaway train, or did those 20 LL test just materialize somehow?


Likewise, 27 entries for exponent 16401349 by user S112725 ("Nick Lipari") appeared in the Sept 21 version of the data files (there were 0 before). After that 29 entries on Sept 29, and 31 entries for Oct 9 and Oct 24.



By the way, the exponent 12178097, mentioned in the first post of this thread, which was LL-tested 26 times by the same user? In the latest version of the data files (Oct 24)... someone found a factor.


Just for fun, I trial-P-1-factored 8893783, no factor found.
GP2 is offline   Reply With Quote
Old 2003-10-26, 17:02   #10
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

32·823 Posts
Default

The Scott Kurowski account is credited with all LL tests that are reported anonymously. This can only be done with the manual web forms. The result line did not contain a "UID: xxx" prefix.
Prime95 is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
colliding trains Fusion_power Puzzles 7 2003-08-30 22:35

All times are UTC. The time now is 00:15.

Fri Apr 16 00:15:49 UTC 2021 up 7 days, 18:56, 1 user, load averages: 1.79, 1.80, 1.83

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

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.