mersenneforum.org  

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

Reply
 
Thread Tools
Old 2021-03-31, 14:48   #1
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

3·5·661 Posts
Default Assignments that are violating the assignment rules

There is a group of exponents in the DC Cat 0 range that are over 90 days old with ZERO progress. They have not checked in since they were assigned. They violate the rules for Cat 1 and Cat 0.
https://www.mersenne.org/assignments...xtf=1&excert=1


This one is even older and has no progress: https://www.mersenne.org/report_expo...5798321&full=1

Most of these P-1 assignments in the Cat 0 range are violating the FTC assignment rules.
https://www.mersenne.org/assignments...xtf=1&excert=1

It would be good to stop handing out FTC's as LL. Any machine that askes for LL FTC's should get DC's.
Uncwilly is online now   Reply With Quote
Old 2021-03-31, 21:10   #2
slandrum
 
Jan 2021
California

2×89 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
There is a group of exponents in the DC Cat 0 range that are over 90 days old with ZERO progress. They have not checked in since they were assigned. They violate the rules for Cat 1 and Cat 0.
https://www.mersenne.org/assignments...xtf=1&excert=1
Not exactly - they weren't cat 0 when they were assigned - the category they were in when assigned is what determines their expiration rules for starting work, even as they move into a different category. The rules for "not starting" are not the same for manual assignments. The "assignment is just too old" rule only kicks in once the assignment has moved 1/2 way into cat 1, and again how old is "too old" is determined by the original category of the assignment.

Any assignment that's over 60 days overdue for reporting in will be expired, but these aren't "overdue", they have next update dates in the future.

The fact that these are manual assignments means they must have been cat 2 (at least) when they were originally assigned.

Last fiddled with by slandrum on 2021-03-31 at 21:24
slandrum is offline   Reply With Quote
Old 2021-04-01, 00:59   #3
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

232738 Posts
Default

Quote:
Originally Posted by Cat 1 DC
Assignments are recycled if assignment is not started within 30 days or does not report in for 30 days or when assignment is more than 60 days old.
If it was Cat 1 when assigned:
(2021-03-30)- (2020-12-24) >> 30 days [greater than 60 days too].

https://www.mersenne.org/report_expo...5798321&full=1 Has had far more than 120 days of a Cat 2 (198 days) I am estimating it was a Cat 2 when it was assigned.

Last fiddled with by Uncwilly on 2021-04-01 at 01:00
Uncwilly is online now   Reply With Quote
Old 2021-04-01, 02:02   #4
slandrum
 
Jan 2021
California

B216 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
If it was Cat 1 when assigned:
(2021-03-30)- (2020-12-24) >> 30 days [greater than 60 days too].

https://www.mersenne.org/report_expo...5798321&full=1 Has had far more than 120 days of a Cat 2 (198 days) I am estimating it was a Cat 2 when it was assigned.
https://www.mersenne.org/M55227059 had to have been cat 2 when it was assigned, it's a manual assignment. That's consistent with the 120 day expiration on it. Since it also had to have been assigned with the "pick lowest exponents" setting on the account to get a manual cat 2 assignment, it would likely have been near to cat 1 when it was assigned.

https://www.mersenne.org/M55798321 was at least cat 3 when it was assigned which would be correct for a 240 day expiration on the assignment. It might have been close to cat 2, but it looks consistent with a cat 3 assignment.
slandrum is offline   Reply With Quote
Old 2021-04-01, 03:25   #5
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

3×5×661 Posts
Default

Quote:
Originally Posted by slandrum View Post
https://www.mersenne.org/M55227059 had to have been cat 2 when it was assigned, it's a manual assignment.
How can you be sure it is a manual assignment?
Uncwilly is online now   Reply With Quote
Old 2021-04-01, 05:31   #6
slandrum
 
Jan 2021
California

2·89 Posts
Default

When I hover over the name of the tester it shows the computer name. In this case it's "Manual Testing" for all of the assignments.
Attached Thumbnails
Click image for larger version

Name:	gimpsreport.jpg
Views:	76
Size:	83.5 KB
ID:	24603  
slandrum is offline   Reply With Quote
Old 2021-04-01, 11:15   #7
S485122
 
S485122's Avatar
 
"Jacob"
Sep 2006
Brussels, Belgium

3×577 Posts
Default

It is possible to view the category boundaries of the assigment rules on a chosen date : one has to append "?dt=yyyy-mm-dd" to the url Assignment rules (https://www.mersenne.org/thresholds).

It shows that the bunch of 39 DC assignments, amongst the lowest Cat 0 exponents and "owned" by user "Christian Geschwentner" were Cat 2 when manually requested ; Assignment rules on 2020-12-24 (https://www.mersenne.org/thresholds/?dt=2020-12-24).

To get rid of this kind of annoyance, a solution could be to limit the time by which one can extend a manual assignment by category. If one wants to have manual assignments in the lower categories, one must herd them. Most users asking manual assignments and working on them don't pile up years, decades or even centuries of work (based on their rate of completion).

Concerning the orphans assignments, they could be cleaned up by applying the first retirement rule :
Quote:
Since PrimeNet began, the server has recycled exponents where the client computer is 60 days past due in updating the server. This rule has been fairly effective and will continue.
One beautiful example showing the rule is not enforced is Active assignments 155000057 : it was last updated in August 2012 when its next update was set 1050 days in the future to June 2015. The rule should have triggered and retired the assignment on the fifth of September 2015, and every day since.The problem concerns all type of assignments, not only LL as in this example.

It would still be possible to get assignments and forget about them while they are worked upon, but the users should check their list regularly if they don't want to loose their fetish numbers ;-)

Jacob

Last fiddled with by S485122 on 2021-04-01 at 11:18 Reason: I still had to correct a spelling mistake after all this editing (there may be more but I don't see them now, maybe later :-(
S485122 is online now   Reply With Quote
Old 2021-04-01, 21:16   #8
drkirkby
 
"David Kirkby"
Jan 2021
Althorne, Essex, UK

1101111112 Posts
Default

Quote:
Originally Posted by S485122 View Post

Concerning the orphans assignments, they could be cleaned up by applying the first retirement rule :
One beautiful example showing the rule is not enforced is Active assignments 155000057 : it was last updated in August 2012 when its next update was set 1050 days in the future to June 2015. The rule should have triggered and retired the assignment on the fifth of September 2015, and every day since.The problem concerns all type of assignments, not only LL as in this example.

It would still be possible to get assignments and forget about them while they are worked upon, but the users should check their list regularly if they don't want to loose their fetish numbers ;-)

Jacob

I've never noticed any that bad before, but I have found it a bit annoying when I wanted to test an exponent and found someone else has got it, but not done anything for over a year.



I started a 100 million digit test some time back. It was the first, and almost certainly the last I attempt


https://www.mersenne.org/report_expo...2646233&full=1


but when I was looking for a 100 million digit exponent, it is annoying to see many smaller ones, which are still 100 million digits, are held by people not working on them.



There is this sentence in the legal section


https://www.mersenne.org/legal/



Terms and Conditions of Use Changes. GIMPS reserves the right to change this TCU without notice. Last updated 15 October 2008.


I think there's an argument for saying that the terms and conditions should be changed. There's a problem that was not foreseen at the time, and the terms and conditions need to be changed to address that problem.



I believe there is a problem, because several people (myself included) seem to get annoyed about assignments that are unavailable for testing, because someone else has held them for a very long time.



Dave
drkirkby is offline   Reply With Quote
Old 2021-05-28, 03:45   #9
Madpoo
Serpentine Vermin Jar
 
Madpoo's Avatar
 
Jul 2014

3,313 Posts
Default

Quote:
Originally Posted by S485122 View Post
...
Concerning the orphans assignments, they could be cleaned up by applying the first retirement rule :
One beautiful example showing the rule is not enforced is Active assignments 155000057 : it was last updated in August 2012 when its next update was set 1050 days in the future to June 2015. The rule should have triggered and retired the assignment on the fifth of September 2015, and every day since.The problem concerns all type of assignments, not only LL as in this example.

It would still be possible to get assignments and forget about them while they are worked upon, but the users should check their list regularly if they don't want to loose their fetish numbers ;-)

Jacob
There has been a longstanding "tradition" that assignments for exponents well beyond the leading edge simply don't have the expiration rules applied. There just wasn't a very compelling reason. There are so many unassigned exponents between A and B that who cares if some particular one is assigned but hasn't been worked on.

What I hear is that some people actually do want particular exponents. In which case, if you determine that it's abandoned, I say go for it. That's not an official sanction of "poaching" assignments, but considering the example you noted, it was last updated 9 years ago and I think it's safe to take it over.

One thing that routinely fascinates me is the degree to which some people really, REALLY, go hard after particular exponents. I see this with TF especially when someone will expend enormous effort factoring a single exponent to ridiculous bit levels (85 and up). Like, why did someone TF M206122907 to 86 bits? Something special about that #?

Or with P-1 factoring to some crazy levels. And I'm not talking about things like M1277 which is currently the lowest unfactored. I get that. Some are goofy, like M1034479 with P1=4,444,444,444 and P2=44,444,444,444

So hey, if you're hung up on a particular exponent, like your psychic told you it is definitely a Mersenne prime, then do what you gotta do.
Madpoo is offline   Reply With Quote
Old 2021-05-28, 15:13   #10
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

3×5×373 Posts
Default

Poaching a Rip Van Winkle exponent is not necessary. A polite request by PM may get the beyond-expired-by-years assignment expired or removed. Just don't be a frequent nuisance about it.
Anyone who wants a challenge is welcome to DC one or more of the hundreds of 100MDigit single-tested exponents, with highly reliable hardware (ECC ram etc) and software, first proving the test environment reliable by more pedestrian 56M-110M DC & PRP without GEC errors. There's around a 19% chance the first test was wrong, judging by the very small sample of ~56 100Mdigit DC/TC tests done so far. (See attachment 3 here.)


(And as Uncwilly subsequently very correctly reminded: first do remedial TF and P-1 to normal-for-the-exponent limits. Not too little, not too much. I generally use the GPU72 row at mersenne.ca as a guide. For example, https://www.mersenne.ca/exponent/332194627 gets a full bounds P-1 run before maybe a DC is run.)

Last fiddled with by kriesel on 2021-05-28 at 16:00
kriesel is offline   Reply With Quote
Old 2021-05-28, 15:26   #11
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

3×5×661 Posts
Default

Quote:
Originally Posted by kriesel View Post
There's around a 19% chance the first test was wrong, judging by the very small sample of ~56 DC/TC tests done so far. (See attachment 3 here.)
That is part of the reason why for years and years, I focused on doing TF in that range. Many of those FTC's on the 100Md numbers were started without enough factoring. The original testers often did zero factoring on them. I tried to eliminate as many exponents as possible factoring with the equipment I had available. I even did some bartering for GPU time in exchange for me doing P-1 on other exponents.
Uncwilly is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
PrimeNet Assignment Rules S485122 PrimeNet 11 2021-05-20 14:54
Modifications to LL assignment rules!!! Prime95 PrimeNet 145 2017-08-05 01:14
Understanding assignment rules Fred PrimeNet 3 2016-05-19 13:40
Tweak to assignment rules Prime95 PrimeNet 11 2014-11-17 02:43
Tweaked assignment rules Prime95 PrimeNet 16 2012-03-19 20:24

All times are UTC. The time now is 17:36.


Sat Sep 18 17:36:21 UTC 2021 up 57 days, 12:05, 0 users, load averages: 3.64, 3.01, 2.45

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.