mersenneforum.org  

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

Reply
 
Thread Tools
Old 2009-08-30, 22:19   #265
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

22·3·599 Posts
Default

Quote:
Originally Posted by storm5510 View Post
This is coming up on six years! What's the deal here?

313?
In addition to what Jacob explained:

As you may already have read in this forum, historically there have not been enough people signing up for double-checks to keep their completions coming in as fast as first-time completions. Thus the gap between the trailing edge of DC results and the trailing edge of LL results has grown fairly steadily over the years.

Quote:
Originally Posted by storm5510 View Post
If a lot this still needs to be done, how does it get assigned?
By signing up for DC assignments from PrimeNet.

If you want to help reduce the six-year gap you noticed, then sign up for DCs and recruit some friends to do so, too.

(I'm running nothing but DCs right now, myself -- though I'll throw in some P-1 for variety once in a while.)

Last fiddled with by cheesehead on 2009-08-30 at 22:36
cheesehead is offline   Reply With Quote
Old 2009-08-30, 22:31   #266
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

1C1416 Posts
Default

Quote:
Originally Posted by Kevin View Post
But still, the fact it takes more than 6 years to get from finding a prime with a first time test to double-checking every exponent beneath it indicates there's a pretty large gap between first time tests and double-checks.
Yep.

Quote:
George has said that he thinks we need more people doing double-checks.
Ditto ... if by "need" one means a desire to see DCs catch up a bit to first-time results.

Quote:
How long is it going to take to prove M(43112609) is the 47th (48th? 49th?) Mersenne prime with the current rate of double-checking
A few years.

Quote:
I mean, "proving" that something actually is the 39th Mersenne prime is kind of an important thing, mathematically, and definitely more important than just clearing all first-time tests beneath a prime.
Okay, but also consider that the search for Mersenne primes goes back hundreds of years (and will continue past our lifetimes). The pace of discovery for the most recent few decades far exceeds the pace before computers.

Take a long-term historical view. It'll get done ... and six years isn't much in the long run. :-)
cheesehead is offline   Reply With Quote
Old 2009-08-30, 22:37   #267
Kevin
 
Kevin's Avatar
 
Aug 2002
Ann Arbor, MI

1B116 Posts
Default

Quote:
Originally Posted by lfm View Post
Yes, you have speculated about such situations before in this thread. Yes, I admit there are no doubt loopholes and opportunities for abuse in the automated system. I think it is unavoidable, something to do with Godel's incompleteness or something (ya I know I am stretching it). I agree fully that we need to watch for abuses but I don't think we can nor should make long lists of rules in advance for things that may never happen. And certainly I would caution against trying to automate such rules.

I think we are basically in agreement here. Keep our eyes open and if something actually happens then we can consider what if anything should be done.
It wouldn't be worth making a long list of rules to try and avoid every possible issue, but I think we might be able to make one or two modifications that prevent the obvious loopholes without infringing on legitimate activity.

For example, say somebody starts off with a C2D, runs 24/7, but they keep a 180 day cache (you can get past the 90 day limit by temporarily removing lines from worktodo.txt, reserving more work, then putting those lines back in). They'd meet the standard of getting preferred exponents pretty quickly, and still be keeping their P4 equivalence up where it needs to be, but ultimately not be returning the assigned preferred exponents for another 6 months. So we could do something where in addition to checking speed/reliability, we could also keep track of ultimate time from assignment to completion and work that into the requirements for preferred exponents. That would only affect people that keep a large cache (past some limit which is more than the Prime95 default maximum of 90 days), and it would only prevent them from getting preferred exponents. I don't think that's especially intrusive.

And FWIW, I checked the two lowest remaining first-time tests, and they've been out for roughly 5 and 9 months, so I don't think my concerns are completely without base. I suppose if in another 3 months the one test is still out, then we can discuss whether or not we think it's a good idea to suggest "pulling the plug" as the assignment system says George would then have the right to do (hopefully he has a super-secret way of seeing how much progress is/has been made on said exponent).
Kevin is offline   Reply With Quote
Old 2009-08-30, 23:03   #268
storm5510
Random Account
 
storm5510's Avatar
 
Aug 2009
U.S.A.

5·337 Posts
Default

Two ideas/suggestions:

1: Cap the work queue to 30 days.

2: If there is no progress reported for more than 90 days, pull everything back.

P95 Options > Preferences has an option for days between reporting expected completion dates. Place a limit of 14 days on it.
storm5510 is offline   Reply With Quote
Old 2009-08-31, 00:03   #269
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

22×3×599 Posts
Default

Quote:
Originally Posted by Kevin View Post
For example, say somebody starts off with a C2D, runs 24/7, but they keep a 180 day cache (you can get past the 90 day limit by temporarily removing lines from worktodo.txt, reserving more work, then putting those lines back in). They'd meet the standard of getting preferred exponents pretty quickly, and still be keeping their P4 equivalence up where it needs to be, but ultimately not be returning the assigned preferred exponents for another 6 months. So we could do something where in addition to checking speed/reliability, we could also keep track of ultimate time from assignment to completion and work that into the requirements for preferred exponents. That would only affect people that keep a large cache (past some limit which is more than the Prime95 default maximum of 90 days), and it would only prevent them from getting preferred exponents.
But what real harm is being done to GIMPS by the situations you decry? I've asked multiple times, but neither you nor anyone else comes up with anything.

Quote:
Originally Posted by storm5510 View Post
1: Cap the work queue to 30 days.
There's no need to. If you think there is a need, please tell us just what real harm is done to GIMPS by allowing the current limits on work queues.

Quote:
2: If there is no progress reported for more than 90 days, pull everything back.
What is the real harm that is being done to GIMPS (not just your own impatience) by allowing PrimeNet and administrators to handle this situation as they already do now?

Quote:
P95 Options > Preferences has an option for days between reporting expected completion dates. Place a limit of 14 days on it.
WHY? What real harm is being done to GIMPS by allowing the current limit?

- - -

You two seem to have a problem with delayed gratification. Can you come up with any real reason other than your inabilities to control your own impatiences for changing the way GIMPS works now?

Last fiddled with by cheesehead on 2009-08-31 at 00:12
cheesehead is offline   Reply With Quote
Old 2009-08-31, 02:00   #270
storm5510
Random Account
 
storm5510's Avatar
 
Aug 2009
U.S.A.

5·337 Posts
Default

Quote:
Originally Posted by cheesehead
You two seem to have a problem with delayed gratification.
Nope, I'm not getting tied up in this mess.

Perhaps, this is why George made the options so broad. Everyone can choose their own preferences.

I'm done here.
storm5510 is offline   Reply With Quote
Old 2009-08-31, 02:42   #271
Kevin
 
Kevin's Avatar
 
Aug 2002
Ann Arbor, MI

433 Posts
Default

Quote:
Originally Posted by cheesehead View Post
But what real harm is being done to GIMPS by the situations you decry? I've asked multiple times, but neither you nor anyone else comes up with anything.
What harm was being done to justify adding the P4 minimum requirements/preferred exponents assignment system? It's precisely the same situation of attempting to prevent the trailing edge from lagging without preventing people from doing the work-type they want.

As you surely know, the situations I've mentioned lead to people attempting to poach exponents (along with other issues which you refuse to accept, so I won't bother mentioning: poaching is quite enough). What harm would come from implementing what I suggested?
Kevin is offline   Reply With Quote
Old 2009-08-31, 03:30   #272
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

212628 Posts
Default

Quote:
Originally Posted by storm5510 View Post
1: Cap the work queue to 30 days.
If most folks just leave the defaults in place when they install, then things will go well. George can dial up or down the cut-offs and speed up the dc rate.
Uncwilly is offline   Reply With Quote
Old 2009-08-31, 03:51   #273
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

587910 Posts
Default

cheesehead uses this argument towards me:
Quote:
Originally Posted by cheesehead View Post
Perhaps you don't value those [credits], but others do.
Kevin uses this argument towards cheesehead:
Quote:
Originally Posted by Kevin View Post
... I think people being disheartened from what appears to be a lack of progress towards goals/milestones/"proving" something is a far more legitimate concern.
But cheesehead refuses to accept Kevin's response and says this:
Quote:
Originally Posted by cheesehead View Post
But what real harm is being done to GIMPS by the situations you decry? I've asked multiple times, but neither you nor anyone else comes up with anything.
cheesehead, you can't have your cake and eat it too. If you argue that some people care about credits (which is fine by itself) then you have to also accept that other people care about the perceived lack of advancement. People care above different things. You could at least acknowledge that others might care about things you don't, since you did argue to me that I should accept that others care about things that I don't. You completely dismissed Kevin's concern as not important. Why would Kevin's concern about advancement of low exponents be any less important than accumulation of credit? Both concerns could potentially drive people away from the project.

cheesehead, you argued that poaching is harmful to the project as a whole by saying that people get upset if their exponent is poached. Why can you not accept that others can get upset if the lower region is not being cleared fast enough? Quid pro quo.
retina is offline   Reply With Quote
Old 2009-08-31, 07:30   #274
davieddy
 
davieddy's Avatar
 
"Lucan"
Dec 2006
England

144638 Posts
Default

http://www.youtube.com/watch?v=TXdk4...eature=related
davieddy is offline   Reply With Quote
Old 2009-08-31, 08:27   #275
Brian-E
 
Brian-E's Avatar
 
"Brian"
Jul 2007
The Netherlands

326510 Posts
Default

Quote:
Originally Posted by Kevin View Post
How do you feel about the ability to have exponents reassigned after one year? Or the fact that there are minimum P4 GHZ requirements for certain worktypes, and the fact that you need to have a certain reliability ranking to get "preferred" exponents? All of those are ways of limiting what slower machines are allowed to do.
...
No problem at all. I am up-front about what my machine capabilities are and how many hours per day it runs. I also don't try to cheat the system by manipulating the worktodo.txt file or anything like that. Therefore I don't expect primenet to allocate me any exponents which would be unsuitable for me and nor do I expect (or ever get!) any hassle for running the client software on my machine for the hours that I indicate I can provide.

And from a different posting of yours:
Quote:
Originally Posted by Kevin View Post
...
And FWIW, I checked the two lowest remaining first-time tests, and they've been out for roughly 5 and 9 months, so I don't think my concerns are completely without base. I suppose if in another 3 months the one test is still out, then we can discuss whether or not we think it's a good idea to suggest "pulling the plug" as the assignment system says George would then have the right to do (hopefully he has a super-secret way of seeing how much progress is/has been made on said exponent).
There's nothing super-secret about it. The regular automatic updates give the percentage from start to completion of the work unit so far. You can see it as part of the line in the prime.log file which reads something like "...&p=44.4750..." . As cheesehead has already indicated, the administrators have the progress-information available on all outstanding work units (except those which are not updating, and these units are canceled and re-assigned after a certain period of tolerance), and they will most certainly have the situation under control as part of the normal running of the project.
Brian-E is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
left shifted S0 value ATH Software 13 2012-09-30 07:19
any mid -level sequence left? firejuggler Aliquot Sequences 5 2012-02-09 11:02
Nothing left to discover? Flatlander Science & Technology 3 2011-09-22 11:19
no twin left behind? Mini-Geek No Prime Left Behind 52 2011-09-12 06:27
New 'No Prime Left Behind' project gd_barnes Lounge 0 2008-01-21 09:05

All times are UTC. The time now is 18:49.

Thu Nov 26 18:49:45 UTC 2020 up 77 days, 16 hrs, 3 users, load averages: 3.65, 2.18, 1.74

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