mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Data (https://www.mersenneforum.org/forumdisplay.php?f=21)
-   -   Reissue exponent to different user/PC? (https://www.mersenneforum.org/showthread.php?t=2078)

markhl 2004-02-12 00:05

Reissue exponent to different user/PC?
 
Does the server automatically issue an exponent to a different PC for double and triple checks? It could check the user/PC requesting work against previous work on that exponent.

It might also be worth issuing an exponent to a different user in case a user has a history of using error-prone PCs.

If it does not already do that, perhaps GP2 or others could review the files to look for cases where 2 matching results are returned by the same user OR the same PC, then we can manually reassign them.

Mitigating factor: as more and more PCs join PrimeNet, it becomes less likely that the same user / PC will get an exponent.

QuintLeo 2004-02-16 09:09

[QUOTE=markhl]Does the server automatically issue an exponent to a different PC for double and triple checks?[/QUOTE]

I don't believe the current server does so automatically, but the probability is fairly low of it happening given the current number of participating machines.

Might be something to mention for the Server V5 project, though.

dswanson 2004-05-12 04:34

I know for certain that the server can and will reissue a doublecheck to the same person turning in the first-time check. I've twice been assigned to doublecheck a number I did the first time around. That's out of over 2000 exponents, which puts the probability around the 0.1% level. And I'm probably an unusual case, since I spent my first few years in GIMPS doing first-time checks, and the past couple doing all doublechecks, which happen to be in the same range first-time checks were a year or two after I started.

robreid 2004-05-12 13:00

[QUOTE=markhl]
If it does not already do that, perhaps GP2 or others could review the files to look for cases where 2 matching results are returned by the same user OR the same PC, then we can manually reassign them.
[/QUOTE]
Within our team I've seen this happen quite regularly, where one team member is assigned to double check an exponent that has already been tested by another team member.

I know Brian Beesley (beejaybee) currently checks for occurances of multiple tests by the same person/team and does a triple check himself.

Here is an example from the The Mersenne Database :
[CODE]
LUCAS_V.TXT(333610): 5893897,Team_Prime_Rib,bayanne_clv1,WX1,B89D507C4EB0870F,..
LUCAS_V.TXT(333611): 5893897,Team_Prime_Rib,garo_shanky,WX1,B89D507C4EB0870F,..
LUCAS_V.TXT(333612): 5893897,beejaybee,nessus-j,WY2,B89D507C4EB0870F,..[/CODE]

dave_0273 2004-05-12 13:30

Some statistics....

There are 297 exponents in lucas_v.txt that have had all of their results returned by the same user id.

Of the 297 mention above, 161 of those exponents have had all their results returned by the same computer id.

(So 136 of the initial 297 had all of its results returned by the same USER ID but by different COMPUTER IDS)

I feel as though the 161 returned by the same computer should be re-run. If I get some time, I will run those computer ids to make sure that none of are considered "error prone". However, there are many computers that have not had a double check on any of their exponents that it is hard to tell if a computer is error prone or not. The only safe way would to run all of these exponents again.

Brian Beesley (beejaybee): Could you please post some information as to how you are going with doublechecking these exponents. I don't want to be stepping on anyone's toes.

garo 2004-05-13 12:51

dave,
Contact Brain Beesley by email or send him a PM. I don't think he checks the forums very often.
Garo


All times are UTC. The time now is 08:58.

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