mersenneforum.org  

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

Reply
 
Thread Tools
Old 2014-11-19, 21:28   #1541
manfred4
 
manfred4's Avatar
 
Mar 2014
Germany

23×3×5 Posts
Default

When every Assignment under the old Rules finally is expired, nobody will have the need of poaching those small exponents anymore. But until then please don't poach the current smallest, without intention I got that one when it was released back 10 days ago :)
The current lowest number will therefore rise again soon...
manfred4 is offline   Reply With Quote
Old 2014-11-19, 21:43   #1542
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

41·227 Posts
Default

Quote:
Originally Posted by manfred4 View Post
But until then please don't poach the current smallest, without intention I got that one when it was released back 10 days ago :)
Are you "NR" by any chance? If so, 53014301 is still yours. I have no intention to "eat it, thinking it were a carrot...".

I also don't generally agree with "poaching" and take such actions carefully, and when possible with general concensus.
chalsall is offline   Reply With Quote
Old 2014-11-19, 23:01   #1543
Primeinator
 
Primeinator's Avatar
 
"Kyle"
Feb 2005
Somewhere near M50..sshh!

2·3·149 Posts
Default

Quote:
Originally Posted by chalsall View Post
I have no intention to "eat it, thinking it were a carrot...".
What if it came with red pepper roasted hummus?
Primeinator is offline   Reply With Quote
Old 2014-11-19, 23:42   #1544
Brian-E
 
Brian-E's Avatar
 
"Brian"
Jul 2007
The Netherlands

2·23·71 Posts
Default

Quote:
Originally Posted by chalsall View Post
As the "poacher", please let me defend myself...

1. From the "instantainious" view from Primenet, these three cadidates /appeared/ to be making progress.

2. From a more temporally spread view of the same report (which I have access to because of my spiders), it was clear that these three candidates would take /much/ more time to actually complete than allowed under the current (implemented) Primenet recycling rules.

3. So, then, I gave notice of my intent to poach, waited for a strong objection, and then loaded them up.



Personally, I'd be very happy with that. I said before that if the straggler turns out to be a MP, the "poached" should be credited.

Edit: Sorry, I misread you. I wouldn't be happy with the result being rejected. I'd be happy with the "poached" being given the credit for the work, even if a MP.
As regards your step number 3, there will be various reasons for those of us who strongly object for not actually speaking up when you wrote that. In my case it boils down to a lack of self confidence related to the fact that I am an extremely tiny contributor to the project, plus a reluctance to spark another tedious argument rehashing lines of discussion which have been repeated ad nauseam.

My favoured approach to your poaching of the last three assignments would have been for PrimeNet to have placed your results on standby without showing them under database enquiries, to have waited for the assignees to finish the work or for the assignments to have expired, and only then to have released your results for general viewing (as DC results in the case of the original assignees finishing the work).

Quote:
Originally Posted by chalsall View Post
[...]I also don't generally agree with "poaching" and take such actions carefully, and when possible with general concensus.
General consensus, unfortunately, is never possible.
Brian-E is offline   Reply With Quote
Old 2014-11-20, 00:15   #1545
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

41·227 Posts
Default

Quote:
Originally Posted by Brian-E View Post
General consensus, unfortunately, is never possible.
Agreed.

This is why some take actions being comfortable within their own skin.
chalsall is offline   Reply With Quote
Old 2014-11-20, 05:15   #1546
ATH
Einyen
 
ATH's Avatar
 
Dec 2003
Denmark

7×419 Posts
Default

Quote:
Originally Posted by Madpoo View Post
Umm... it bears mentioning that all 3 of those were poached.

One of them, 51907363, was making steady progress and being updated daily with an ETA of Dec 1. The other 2 in the 52M range had last checked in 6 days ago, and were 81% done, with ETAs of Dec 10 and 11, so they weren't really abandoned either.
Quote:
Originally Posted by lycorn View Post
I was quite displeased by the poaching of these 3 exponents, that were making a steady progress and approaching completion at a regular pace. I really don´t understand the motivation for doing this in such circumstances, apart from an unjustified impatience, or some desire of being noticed.
Did you both miss the discussion we had 5-7 days ago in this thread about those 3 exponents?

I do not understand why primenet and you would think that if an exponent has taken 576 days to reach 81% (0.140625% per day) why would it suddenly complete the last 19% in just ~ 22 days (0.863636% per day). Those ETA's makes no sense. George posted the SQL code for recycling and 51907363 was due for recycling Nov 19th or 20th, and the 2 others around ~ Dec 9th. Also Chris had his own statistics that showed all 3 would not be complete for ages:
http://www.mersenneforum.org/showpos...postcount=1494
ATH is online now   Reply With Quote
Old 2014-11-20, 07:50   #1547
lycorn
 
lycorn's Avatar
 
Sep 2002
Oeiras, Portugal

1,399 Posts
Default

I missed those details, indeed.
Although I appreciate that the matter was pondered, so it wasn´t just an "impulse", I still advocate that we should abide by the rules we agreed as a community regarding the recycling of exponents, even though sometimes it really seems like we are not doing any harm in poaching.
That said, I owe an apology to chalsall: in the light of the post quoted by ATH I acknowledge that "an unjustified impatience, or some desire of being noticed" was a bit too strong. No hard feelings?
lycorn is offline   Reply With Quote
Old 2014-11-20, 09:23   #1548
ATH
Einyen
 
ATH's Avatar
 
Dec 2003
Denmark

7·419 Posts
Default

Can the ETA calculation for those very slow exponent be improved? Maybe based on the average speed during the last say 3 reports to the server?:

ETA date = date of last report to server + (100% - last report progress)/("amount of progress since 3rd-last report"/days since 3rd last report)



For example for 52957519 and 52983583 they had according to Chris ~10% progress since 2014-05-17, so lets say that was their 3rd last report, and last report was roughly Nov 13th at 81%:

ETA date = 2014-11-13 + (100%-81%) / (10%/180 days) = 2015 Oct 21st.

Last fiddled with by ATH on 2014-11-20 at 09:38
ATH is online now   Reply With Quote
Old 2014-11-20, 13:31   #1549
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

41·227 Posts
Default

Quote:
Originally Posted by lycorn View Post
No hard feelings?
None at all; no problem.
chalsall is offline   Reply With Quote
Old 2014-11-21, 05:27   #1550
Madpoo
Serpentine Vermin Jar
 
Madpoo's Avatar
 
Jul 2014

2·1,637 Posts
Default

Quote:
Originally Posted by ATH View Post
I do not understand why primenet and you would think that if an exponent has taken 576 days to reach 81% (0.140625% per day) why would it suddenly complete the last 19% in just ~ 22 days (0.863636% per day). Those ETA's makes no sense.
They may have been assigned, but behind some other work on the same machine. I forget what the default and max settings are for how much work can be queued up in the worktodo file.

George had also mentioned that there is a limit on the rolling average of "500" (50% of the base 1000) or whatever... well, I may have mangled that, but I understood that essentially there's a low end value on that where it assumes a machine will at least be on 12 hours a day running at peak efficiency. If a machine is only turned on and running P95 maybe 2-4 hours a day, its estimates are going to be wildly off because it's still assuming 12 hours of work per day.

Or maybe some folks could be running P95 on a heavily burdened system, and P95 is only getting minimal CPU cycles, so the same lower threshold being reached and breached is resulting in those inaccurate estimates.

The end result, I guess, is that each time it reports in (which is daily in many cases), it reports very small progress from the day before, and keeps moving the ETA further and further out. At first glance it would seem weird and suspicious, like... "what is that person trying to pull?" but I think it's nothing more than a machine running at < 50% efficiency... sometimes probably by quite a bit.

A machine where the rolling average is more like 25% is going to be off by twice as many days as it's original estimate... and since the "days of work to queue" is also basing it's decision to return work that is out too far on that rolling average, it could very well result in too much work being queued up than it really ought to.

So... that's my speculation on how/why it happens.

The grandfathered exponents will run their course before too long though. It's more pronounced with those assignments made prior to 2014-03-01. I think the assignments made after are less tolerant of that kind of accidental tomfoolery. The lack of progress after so many days from being assigned will recycle it MUCH sooner. Clients that don't pay attention could still start working on it at some point, but that's kind of their problem, ya know? Or maybe the client will see an expired assignment when it checks in, and if work hasn't begun it gets removed automatically?
Madpoo is offline   Reply With Quote
Old 2014-11-21, 06:12   #1551
Madpoo
Serpentine Vermin Jar
 
Madpoo's Avatar
 
Jul 2014

2×1,637 Posts
Default Reddit on the M44 announcement

I ran across this link when looking at traffic to the site recently:
http://www.reddit.com/r/math/comment...ersenne_prime/

I thought it was mildly interesting (and a little amusing at some of the misunderstandings about what the announcement really was saying).
Madpoo is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Newer X64 build needed Googulator Msieve 73 2020-08-30 07:47
Performance of cuda-ecm on newer hardware? fivemack GMP-ECM 14 2015-02-12 20:10
Cause this don't belong in the milestone thread bcp19 Data 30 2012-09-08 15:09
Newer msieves are slow on Core i7 mklasson Msieve 9 2009-02-18 12:58
Use of large memory pages possible with newer linux kernels Dresdenboy Software 3 2003-12-08 14:47

All times are UTC. The time now is 04:57.

Tue Sep 22 04:57:28 UTC 2020 up 12 days, 2:08, 0 users, load averages: 1.58, 1.61, 1.54

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.