Assignment Problems

Prime95 · 2012-03-08, 00:14

A manual user emailed me that these exponents are now registered to GPUto72:

50311747 (running, non-networked)
42576463 (running, networked)
45522661 (queued on networked CPU)
42578279 (ditto)
43094647 (running, non-networked)
43163753 (queued on networked CPU)
42578609 (running on networked CPU)

The server should not have handed these out - so some investigation is required. In the meantime, can these exponents be quarantined and not handed out to any users (or if already handed out can you please stop testing until the user has a chance to complete his first time test)?

Dubslow · 2012-03-08, 00:39

Huh. chalsall, what exactly did you teach Spidey?

@George, here's a link to what is publicly known. You'll need chalsall to get more details.
http://www.mersenneforum.org/showthr...=16263&page=32

Note that Spidey claimed 20 expos below M47, with reserve dates occasionally a year or more before GPU272 was even created; these low expos were passed out to forum-goers for "manual testing", meaning Spidey did not unreserve them and reclaim them as ANON as is done for 45M tests.

I have at least one of those expos listed above, as well as two others that were obtained as above -- would you like me to PM you which ones? Of course, chalsall can tell you all of them in one go.

I will fill up on P-1 in the mean time; I'm at about 57% on the expo in that list, we can trade save files if the owner wants.

kladner · 2012-03-08, 01:05

Whew! At first look I thought I had one of those, at 36.7% completion. A closer look showed that just the last three digits were the same (661).

Dubslow · 2012-03-08, 02:55

For example, for the exponent I have on that list, PrimeNet thinks that GPU272 has had the assignment for 1200 days. (Again, can PM as necessary.)

chalsall · 2012-03-08, 17:54

Quote:

Originally Posted by Prime95

A manual user emailed me that these exponents are now registered to GPUto72:

50311747 (running, non-networked)
42576463 (running, networked)
45522661 (queued on networked CPU)
42578279 (ditto)
43094647 (running, non-networked)
43163753 (queued on networked CPU)
42578609 (running on networked CPU)

The server should not have handed these out - so some investigation is required. In the meantime, can these exponents be quarantined and not handed out to any users (or if already handed out can you please stop testing until the user has a chance to complete his first time test)?

I hope people don't think I'm ignoring this; I'm not. But I have "real" work to do as well, and it took a bit of time to drill down on these candidates against the knowledge I happen to have.

George... I used a new technique to aquire candidates, which I will PM you the details about shortly. It was, in my opinion, entirely "legal", and I used a "slaved" mprime and the PrimeNet API to do so. As in, it was entirely your code which did this.

However, I think four things are important...

1. Please note the reservation dates of the above candidates:

Code:

42576463 - 2008-11-24
42578279 - 2008-11-30
42578609 - 2008-12-08
43094647 - 2008-12-02
43163753 - 2008-12-02
45522661 - 2008-11-25
50311747 - 2008-10-28

1.1. Please note that no LL work had been done on them (or, at least, not reported).

2. The user in question is not questioning the "loss" of approximately 200 other candidates.

3. The user in question produces approximately two (2) LL results a month.

4. The user in question also has four "Trial Factoring" assignments from 2008-11-26, 1191 days ago, which have been independently completed by other workers.

4.1. To be explicit, the TF work was not coordinated by GPU72.

4.2. To be more explicit, the candidates I'm talking about are 51130021, 51163129, 51208343 and 51208343.

4.2.1. Don't you think it might make sense for PrimeNet to release these back into the Pool for LL work?

At the end of the day the GPU to 72 project's goal is to help find the next Mersenne Prime.

If finding bugs in PrimeNet is part of the process, I would argue that is simply a healthy example of the Scientific Method.

chalsall · 2012-03-09, 00:56

So, as an experiment, I manually requested "Trial Factoring" work from PrimeNet shortly after midnight UTC.

I was given two assignments below 45M to TF from 69 to 70.

Note that these are below where "Spidy" is given assignments. And further note that these were assigned to "Anonymous" to LL on "2011-07-20". According to the historical data I have, these were both at approximately 43% complete of the LL test.

George -- should I "hold off" doing any testing on these as well?

Prime95 · 2012-03-09, 02:24

Quote:

Originally Posted by chalsall

So, as an experiment, I manually requested "Trial Factoring" work from PrimeNet shortly after midnight UTC.

I was given two assignments below 45M to TF from 69 to 70.

Note that these are below where "Spidy" is given assignments. And further note that these were assigned to "Anonymous" to LL on "2011-07-20". According to the historical data I have, these were both at approximately 43% complete of the LL test.

George -- should I "hold off" doing any testing on these as well?

I don't have strong feelings one way or the other on "holding off". I need to drill down and figure out why Primenet is releasing these as TF assignments.

To be clear, I'm not accusing anyone of doing anything underhanded. I just need to figure out how it happened and correct it. The user in question has been around a long time contributing with slow machines. He obviously becomes frustrated when "poached".

As to recycling seemingly abandoned TF assignments, we need to look into this. It is a separate issue.

Prime95 · 2012-03-10, 21:41

The user also reports these exponents were scooped:

42578197
42652487
42968837
43025869
43055297
43110383

If you have one of them, please don't LL it.

Chalsall and I think we have figured out why this has happened so we can prevent this in the future.

Dubslow · 2012-03-10, 21:51

This is perhaps a bold proposition, but it is based on chalsall's observation that this user only completes two LLs per month. From the initial post here, it seems about half of the expos are not yet running. Seeing as these are milestone exponents, and we (you) are able to communicate with said user, would he be willing to not run those which aren't already going, and instead pass them out to forum goers here so that all those low expos can be run "in parallel" and completed as soon as possible? The user can easily reserve more (higher?) exponents from PrimeNet. I make this suggestion because the expo I got on the initial list in this thread would be done roughly tomorrow if I hadn't paused it. (I will finish the DC when this user finishes his first run.) Of course, if the user wants to keep the low exponents and do them himself (or herself) then that's obviously fine.

(Implementation wise, this user would keep the assignment ID from PrimeNet, but others here run the tests.)

Prime95 · 2012-03-10, 22:13

Quote:

Originally Posted by Dubslow

This is perhaps a bold proposition...

I've just emailed the user explaining what happened. The fairest thing is to let the user reclaim as many of his exponents as possible. The fact that they are "milestone exponents" is irrelevant.

I've also suggested he join gpu72.com to get LL tests in the future. He enjoys testing smaller exponents (as do I) and this is the best way to get them.

chalsall · 2012-03-10, 22:15

Quote:

Originally Posted by Dubslow

From the initial post here, it seems about half of the expos are not yet running. Seeing as these are milestone exponents, and we (you) are able to communicate with said user, would he be willing to not run those which aren't already going, and instead pass them out to forum goers here so that all those low expos can be run "in parallel" and completed as soon as possible?

I'm glad someone said it...

I have to say I agree. Why should someone be able to reserve (several hundred) candidates, and then sit on them for more than three years?

Please note the following data on these latest six candidates:

Code:

Candidate - Reserved Date - Estimated Completion
42578197 - 2008-12-02 - 2009-02-08
42652487 - 2008-11-24 - 2011-02-22
42968837 - 2008-12-07 - 2009-04-19
43025869 - 2008-11-24 - 2011-03-01
43055297 - 2008-11-24 - 2011-03-01
43110383 - 2008-11-24 - 2011-03-02

George... While I will abide by anything you tell me to do with regards to PrimeNet and this project, it has to be said that all of these would have been recycled long ago if PrimeNet followed the expiry rules you yourself have said are in place.

2012-03-08, 00:14	#1
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 7526₁₀ Posts	Assignment Problems A manual user emailed me that these exponents are now registered to GPUto72: 50311747 (running, non-networked) 42576463 (running, networked) 45522661 (queued on networked CPU) 42578279 (ditto) 43094647 (running, non-networked) 43163753 (queued on networked CPU) 42578609 (running on networked CPU) The server should not have handed these out - so some investigation is required. In the meantime, can these exponents be quarantined and not handed out to any users (or if already handed out can you please stop testing until the user has a chance to complete his first time test)?

2012-03-08, 00:39	#2
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 1C35₁₆ Posts	Huh. chalsall, what exactly did you teach Spidey? @George, here's a link to what is publicly known. You'll need chalsall to get more details. http://www.mersenneforum.org/showthr...=16263&page=32 Note that Spidey claimed 20 expos below M47, with reserve dates occasionally a year or more before GPU272 was even created; these low expos were passed out to forum-goers for "manual testing", meaning Spidey did not unreserve them and reclaim them as ANON as is done for 45M tests. I have at least one of those expos listed above, as well as two others that were obtained as above -- would you like me to PM you which ones? Of course, chalsall can tell you all of them in one go. I will fill up on P-1 in the mean time; I'm at about 57% on the expo in that list, we can trade save files if the owner wants. Last fiddled with by Dubslow on 2012-03-08 at 00:43

2012-03-10, 21:51	#9
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	This is perhaps a bold proposition, but it is based on chalsall's observation that this user only completes two LLs per month. From the initial post here, it seems about half of the expos are not yet running. Seeing as these are milestone exponents, and we (you) are able to communicate with said user, would he be willing to not run those which aren't already going, and instead pass them out to forum goers here so that all those low expos can be run "in parallel" and completed as soon as possible? The user can easily reserve more (higher?) exponents from PrimeNet. I make this suggestion because the expo I got on the initial list in this thread would be done roughly tomorrow if I hadn't paused it. (I will finish the DC when this user finishes his first run.) Of course, if the user wants to keep the low exponents and do them himself (or herself) then that's obviously fine. (Implementation wise, this user would keep the assignment ID from PrimeNet, but others here run the tests.) Last fiddled with by Dubslow on 2012-03-10 at 21:53 Reason: grammar

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
PC problems	Nimras	Information & Answers	6	2009-12-15 21:24
Need help with few problems	Laserjet	Hardware	1	2007-10-13 10:59
Two problems	gribozavr	Puzzles	11	2007-02-05 05:46
Sieving Problems	amcfarlane	Miscellaneous Math	5	2006-08-01 23:31
Heat problems	michaf	Sierpinski/Riesel Base 5	3	2006-07-04 21:11

2012-03-08, 01:05	#3
kladner "Kieren" Jul 2011 In My Own Galaxy! 2×3×1,693 Posts	Whew! At first look I thought I had one of those, at 36.7% completion. A closer look showed that just the last three digits were the same (661).

2012-03-08, 02:55	#4
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 7221₁₀ Posts	For example, for the exponent I have on that list, PrimeNet thinks that GPU272 has had the assignment for 1200 days. (Again, can PM as necessary.)

2012-03-09, 00:56	#6
chalsall If I May "Chris Halsall" Sep 2002 Barbados 2·5·7·139 Posts	So, as an experiment, I manually requested "Trial Factoring" work from PrimeNet shortly after midnight UTC. I was given two assignments below 45M to TF from 69 to 70. Note that these are below where "Spidy" is given assignments. And further note that these were assigned to "Anonymous" to LL on "2011-07-20". According to the historical data I have, these were both at approximately 43% complete of the LL test. George -- should I "hold off" doing any testing on these as well?

2012-03-10, 21:41	#8
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 2×53×71 Posts	The user also reports these exponents were scooped: 42578197 42652487 42968837 43025869 43055297 43110383 If you have one of them, please don't LL it. Chalsall and I think we have figured out why this has happened so we can prevent this in the future.