mersenneforum.org  

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

Reply
 
Thread Tools
Old 2015-06-09, 15:20   #1
CuriousKit
 
"J. Gareth Moreton"
Feb 2015
Nomadic

2×32×5 Posts
Default Unsafe verification?

I was just browsing around the Recent Cleared list out of curiosity and found a couple of anomalies. One is that a lot of the factors are truncated to the first 14 digits, which led to some brief confusion because a lot of them appeared to be even numbers, although a closer inspection cleared that up... it's still something that should possibly be addressed though.

But one thing I stumbled upon is the verification for M54357769:

The two completion dates are only a day apart. It could be just a coincidence and two people got the same LL assignment simultaneously somehow (and if that's true, then the result's genuine, and it looks like that MadPoo's was run on a GPU or some other manual testing framework), but it feels a bit shaky at best, especially as one of the tests probably hasn't been run with the double-check safeguards.

Is there a way to manually acquire a double-check job for a number that has already been marked as cleared? The manual testing assignment page just gave me an error.
CuriousKit is offline   Reply With Quote
Old 2015-06-09, 15:35   #2
Mini-Geek
Account Deleted
 
Mini-Geek's Avatar
 
"Tim Sorbera"
Aug 2006
San Antonio, TX USA

102538 Posts
Default

Quote:
Originally Posted by CuriousKit View Post
I was just browsing around the Recent Cleared list out of curiosity and found a couple of anomalies. One is that a lot of the factors are truncated to the first 14 digits, which led to some brief confusion because a lot of them appeared to be even numbers, although a closer inspection cleared that up... it's still something that should possibly be addressed though.
Looks like the table is wider than its containing element (thanks in large part to M1,785,851's enormous 55-digit composite P-1 factor), so you have to scroll to see it all. Do note, though, that the left-aligned things with 14 (hex)digits and then __ are not factors, but masked LL residues. Those can legitimately be even (or look even with their masks). It's the right-aligned things with (only decimal) digits and a varying number of digits that are factors (those might only look even because of the scrolling).

Quote:
Originally Posted by CuriousKit View Post
But one thing I stumbled upon is the verification for M54357769:

The two completion dates are only a day apart. It could be just a coincidence and two people got the same LL assignment simultaneously somehow (and if that's true, then the result's genuine, and it looks like that MadPoo's was run on a GPU or some other manual testing framework), but it feels a bit shaky at best, especially as one of the tests probably hasn't been run with the double-check safeguards.

Is there a way to manually acquire a double-check job for a number that has already been marked as cleared? The manual testing assignment page just gave me an error.
A quick google shows that Madpoo was closely watching the first user's progress, ran it earlier himself (without a PrimeNet reservation), and reported his result soon after the first was reported. As for the safeguards, every test run by Prime95 has them (i.e. a randomized starting point that doesn't affect the end result, IIRC), regardless of whether it thinks it's a DC. The only concern would be if both LLs were run on GPUs (which doesn't appear to be the case here).

Last fiddled with by Mini-Geek on 2015-06-09 at 15:41
Mini-Geek is offline   Reply With Quote
Old 2015-06-09, 15:44   #3
CuriousKit
 
"J. Gareth Moreton"
Feb 2015
Nomadic

2·32·5 Posts
Default

Oh right, okay. Thanks. It just seemed a little strange at first glance, but as long as the residues as good, then there's no worry.

As for the truncated values, it was decimal factors I noticed - for example:

E.Lechner - 960334201 - F - 2015-06-09 14:47 - 1.5 - 0.0018 - 13173418398624 (The actual factor is 131734183986244159207)


The length of truncation feels a little arbitrary, although truncating ones with a truly massive length makes sense, especially if the factor is composite, as with your example or, say M1722131 (displayed as 39969646292585377336075974182308).

Last fiddled with by CuriousKit on 2015-06-09 at 15:55
CuriousKit is offline   Reply With Quote
Old 2015-06-09, 16:04   #4
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

10010011111002 Posts
Default

Might be your browser....when I click on the links in your post I see ALL digits of the factors

Quote:
Mn Status Details
1722131 Factored 413311441
4747402283378392993
84192667709090339551
History
Date User Type Result
2015-06-08 alpertron F-PM1 Factor: 399696462925853773360759741823089166143 / (P-1, B1=500000, B2=15000000, E=12)
2015-06-08 alpertron F-PM1 Factor: 399696462925853773360759741823089166143 / (P-1, B1=500000, B2=15000000, E=12)

Last fiddled with by petrw1 on 2015-06-09 at 16:04
petrw1 is offline   Reply With Quote
Old 2015-06-09, 16:06   #5
CuriousKit
 
"J. Gareth Moreton"
Feb 2015
Nomadic

2×32×5 Posts
Default

The numbers are displayed in full when you click those links, yeah. I was meaning the "Recent Cleared" list (sort by descending order of exponent to see a clear case of multiple truncated factors).
CuriousKit is offline   Reply With Quote
Old 2015-06-09, 16:28   #6
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

22·7·132 Posts
Default

Quote:
Originally Posted by CuriousKit View Post
The numbers are displayed in full when you click those links, yeah. I was meaning the "Recent Cleared" list (sort by descending order of exponent to see a clear case of multiple truncated factors).
This is what mine shows: ... not truncated.

A copy/paste from: http://www.mersenne.org/report_recent_cleared/


Code:
William Wallace		77261363	F-PM1	2015-06-08 17:16	6.0	2.9231	652570191930242911372515684959
alpertron	Manual testing	1769441	F-PM1	2015-06-06 22:27	0.0	0.1015	2164970057879540403248790630348635207
alpertron	Manual testing	1769441	F-PM1	2015-06-06 22:27	0.0	0.1015	2164970057879540403248790630348635207
alpertron	Manual testing	1722131	F-PM1	2015-06-08 11:24	0.0	0.1015	399696462925853773360759741823089166143
alpertron	Manual testing	1722131	F-PM1	2015-06-08 11:24	0.0	0.1015	399696462925853773360759741823089166143
TheJudger	Manual testing	74153621	F-PM1	2015-06-07 20:30	0.0	2.8004	142294638775192859584174293669726827722910365592351
TheJudger	Manual testing	74153621	F-PM1	2015-06-07 20:30	0.0	2.8004	142294638775192859584174293669726827722910365592351
mikr	MSI	1785851	F-PM1	2015-06-08 18:50	0.0	0.5555	8044011922388623641167074992834902439100929361497037199

and a screen print (cropped) below
Attached Thumbnails
Click image for larger version

Name:	screen print.png
Views:	134
Size:	35.3 KB
ID:	12724  

Last fiddled with by petrw1 on 2015-06-09 at 16:31 Reason: added screen print
petrw1 is offline   Reply With Quote
Old 2015-06-09, 16:45   #7
CuriousKit
 
"J. Gareth Moreton"
Feb 2015
Nomadic

10110102 Posts
Default

How do these ones appear to you? Apologies, I don't mean to raise an argument - it just seems to be a little inconsistent to me.
Attached Thumbnails
Click image for larger version

Name:	Truncated.png
Views:	151
Size:	99.4 KB
ID:	12725  
CuriousKit is offline   Reply With Quote
Old 2015-06-09, 16:49   #8
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

160658 Posts
Default

Quote:
Originally Posted by CuriousKit View Post
How do these ones appear to you? Apologies, I don't mean to raise an argument - it just seems to be a little inconsistent to me.
Is there a horizontal scroll bar at the bottom of your table? The right side appears to be missing its border, which the left side clearly has.
Dubslow is offline   Reply With Quote
Old 2015-06-09, 16:55   #9
CuriousKit
 
"J. Gareth Moreton"
Feb 2015
Nomadic

2·32·5 Posts
Default

*facepalms* Yes there is... right at the bottom of the very long list.

Can this topic be deleted now? I need to go away in shame!
CuriousKit is offline   Reply With Quote
Old 2015-06-10, 02:16   #10
Madpoo
Serpentine Vermin Jar
 
Madpoo's Avatar
 
Jul 2014

1100111100012 Posts
Default

Quote:
Originally Posted by CuriousKit View Post
*facepalms* Yes there is... right at the bottom of the very long list.

Can this topic be deleted now? I need to go away in shame!
We forgive you. It could be my fault anyway. That "recent cleared" report (and all the pages, really) have a fixed-width of 1000 pixels. You might be rocking a screen narrower than that? Or you're not running your browser window maximized?

Maybe in some future "round two" of web redesign I could tackle making the pages "responsive". That's the big thing these days and actually factors in to how Google ranks your website. At my day job we recently did all that so the page will resize, collapse menus, even remove menu options and hide them away, if you make your browser window narrower.

Gone are the days when some websites might have separate desktop and mobile versions, the latter having some features removed or simplified... no... "responsive" is the buzzword these days. And it never ends... it's hard enough to get a site running okay on multiple browsers (Chrome, IE, Safari), and adding in responsive web design apparently ratchets it up several notches. You're dealing with mobile versions of popular browsers which, surprise, act different than their desktop counterpart.

I don't do the web design, development, etc. but I hear the conversations from the people that do, and honestly, I'd pull my hair out if I were them. They're made of tougher stuff than I.

Oh, and yes, I checked in that other result for M54357769 ... read the "milestones" thread for the latest juicy gossip on how we've been tracking the progress of a few assignments in particular.
Madpoo is offline   Reply With Quote
Old 2015-06-10, 02:21   #11
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
Jun 2011
Thailand

261816 Posts
Default

Quote:
Originally Posted by CuriousKit View Post
*facepalms* Yes there is... right at the bottom of the very long list.

Can this topic be deleted now? I need to go away in shame!
No need to delete the topic. This is a known problem and some of us are pissed off with it, especially when you display the lifetime tops for "all type" of the work, and you need to scroll down the page (well, the "home" and "end" keys on the keyboard help, please use them, but still is pissing me off!) to be able to see the last 3 digits of the "overtake" column. We tried to convince Madpoo to give us few more digits on the right of that table but it was not an important thing on his list (well, we are convinced that there are other things more important, but it is still pissing us off). A simple fix would be to double the horizontal scrollbar with another one in the top.

We even went so far as to create a firefox template for this page, loaded every time we access this top, which essentially changes the width of the white area from 99.4% to 102%, and we can see all the digits on the right (you can do this in ff, right click on the page and choose "inspect element").
Click image for larger version

Name:	percent.PNG
Views:	136
Size:	137.8 KB
ID:	12730

[edit: (crosspost)
Ha! you are here on the barricade! Happy to see you! Maybe you hear me this time... hehe]

[edit 2: about that exponent being DC in about a day, even without "gossips" we did in the parallel thread, there are GPU cards around which can do the job in less than 30 hours, so it would be no wonder...]

Last fiddled with by LaurV on 2015-06-10 at 02:30
LaurV is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
mfaktc: potentially unsafe inline asm Fredrik GPU Computing 9 2016-01-15 22:29
Doing more about LL test verification jasong Software 13 2010-03-10 18:53
linux 25.9 client, unsafe thread output? xorbe Software 0 2009-04-03 04:21

All times are UTC. The time now is 07:08.


Sat Sep 18 07:08:59 UTC 2021 up 57 days, 1:37, 0 users, load averages: 0.98, 1.13, 1.23

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.