mersenneforum.org  

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

Reply
 
Thread Tools
Old 2019-04-05, 20:24   #1
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

23×7×167 Posts
Default Crazy spike in LL tests submitted

The graph here: https://www.mersenne.org/primenet/graphs.php shows a doubling of first time LL submissions.

Aaron is something fishy going on?
Uncwilly is online now   Reply With Quote
Old 2019-04-06, 22:13   #2
Madpoo
Serpentine Vermin Jar
 
Madpoo's Avatar
 
Jul 2014

329010 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
The graph here: https://www.mersenne.org/primenet/graphs.php shows a doubling of first time LL submissions.

Aaron is something fishy going on?
I'm checking it out. There were 252 LL results that came from a single user (public name is ANONYMOUS but it's an actual registered user in there).

The exponents were all assigned in February between the 11th and 15th and were then all checked in using the manual submission page on April 4.

The result format is something I'm not familiar with... It's not a Prime95 client, unless it's something ancient which should raise a flag.

Anyone familiar with a client that spits out LL results like this (I masked the residue):
M89224181 is not prime. Res64: 9B8E9EA612CCFB__. Program: E18.0. Final residue shift count = 87396767

For someone to reserve a bunch and then check in 252 in a single day, less than 2 months since being assigned (really more like 7 weeks) is... unusual.

They did also have one result that came in back on Feb 25th, 2 weeks after it was assigned on the 11th along with a bunch of others: M89224181

So it seems weird, but on the other hand, there are some details about the user registration (non-public) that make me think these could be legit. I just don't know what that result string means... what is "Program: E18.0" ? I'll have to research that or figure it out before my gut leads me one way or another.

EDIT: I see that "Program: E18.0" is mlucas. Maybe we should ask Ernst if he has anyone doing testing on a swarm of machines.

Last fiddled with by Madpoo on 2019-04-06 at 22:15
Madpoo is offline   Reply With Quote
Old 2019-04-06, 23:27   #3
lycorn
 
lycorn's Avatar
 
Sep 2002
Oeiras, Portugal

26508 Posts
Default

What about putting some of them, randomly chosen, to an early double check?
lycorn is offline   Reply With Quote
Old 2019-04-07, 03:02   #4
Madpoo
Serpentine Vermin Jar
 
Madpoo's Avatar
 
Jul 2014

63328 Posts
Default

Quote:
Originally Posted by lycorn View Post
What about putting some of them, randomly chosen, to an early double check?
If anyone is game, first 10 exponents (by size):
89223847
89223917
89223991
89224049
89224073
89224141
89224147
89224181
89247787
89247799
Madpoo is offline   Reply With Quote
Old 2019-04-07, 03:13   #5
GP2
 
GP2's Avatar
 
Sep 2003

1010000101112 Posts
Default

Quote:
Originally Posted by Madpoo View Post
I already took this one. Should finish in three days.
GP2 is offline   Reply With Quote
Old 2019-04-07, 17:39   #6
Madpoo
Serpentine Vermin Jar
 
Madpoo's Avatar
 
Jul 2014

2·5·7·47 Posts
Default

Quote:
Originally Posted by GP2 View Post
I already took this one. Should finish in three days.
Let me just say that I have a theory this could have something to do with Ernst's "cell phone cluster" discussion going on in another thread.
Madpoo is offline   Reply With Quote
Old 2019-04-07, 19:34   #7
ewmayer
2ω=0
 
ewmayer's Avatar
 
Sep 2002
República de California

2×7×829 Posts
Default

Quote:
Originally Posted by Madpoo View Post
Let me just say that I have a theory this could have something to do with Ernst's "cell phone cluster" discussion going on in another thread.
You're close - same client (ARMv8 SIMD build), but the 250+ 1st-time checks are from a "power user" I've been trading e-mails with re. code issues (he was the fellow who alerted me to several instances of missing register names in GCC asm-macro clobber lists for the ARMv8 asm macros that first appeard in v17.1, those incomplete clobber lists appear to have been responsible for the "bad build" issues with the v17.1 client on ARM) for the past several months, who had first-look access to a bunch of manycore (224-256 cores per box) ARM-based servers in a data center he appears to be on the IT squad for. The servers are now loaded with higher-priority work most of the time, but he apparently had almost 100% use for on the order of a month. IOW the results are legit, but I support the idea of subjecting a random sample to immediate DCs.

My own cellphone cluster is about half built out, that will end up with a dozen Galaxy S7 broke-o-phones, and - after each node has successfully done a couple DCs to verify hardware integrity in 24/7 all-cores-crunching mode - should average ~1 first-time test every 7-10 days. If after completing the proof-of-principle groundwork I can convince a couple hundred linux-enthusiast hobbyists around the world to build similar desktop mini-clusters, we can make a similar project-throughput-boosting contribution in a more sustainable and scalable manner. You can think of the above user as the Curtis Cooper of my "ARMy ant distributed computation effort". :)

Last fiddled with by ewmayer on 2019-04-07 at 19:39
ewmayer is offline   Reply With Quote
Old 2019-04-07, 21:35   #8
ixfd64
Bemusing Prompter
 
ixfd64's Avatar
 
"Danny"
Dec 2002
California

22·3·197 Posts
Default

I've always been a little envious of IT folks for having access to so much computing power. It's not surprising IT professionals make up a considerable percentage of Mersenne prime discoverers.

Last fiddled with by ixfd64 on 2019-04-07 at 21:35
ixfd64 is offline   Reply With Quote
Old 2019-04-08, 02:20   #9
ewmayer
2ω=0
 
ewmayer's Avatar
 
Sep 2002
República de California

2×7×829 Posts
Default

Quote:
Originally Posted by ixfd64 View Post
I've always been a little envious of IT folks for having access to so much computing power. It's not surprising IT professionals make up a considerable percentage of Mersenne prime discoverers.
I believe our own MadPoo has a bit of personal experience on that front. :)
ewmayer is offline   Reply With Quote
Old 2019-04-08, 02:37   #10
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

23×7×167 Posts
Default

Quote:
Originally Posted by ewmayer View Post
I believe our own MadPoo has a bit of personal experience on that front. :)

If I could get all of my employer's machines running Prime95 I would be in the top 40 or 20.
Uncwilly is online now   Reply With Quote
Old 2019-04-08, 15:15   #11
lukerichards
 
lukerichards's Avatar
 
"Luke Richards"
Jan 2018
Birmingham, UK

25·32 Posts
Default

Quote:
Originally Posted by Uncwilly View Post

If I could get all of my employer's machines running Prime95 I would be in the top 40 or 20.
My work has 100+ quad core (4 physical cores) machines plus an RDP server with God knows how many CPUs.

If only...
lukerichards is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Manually submitted P-1 results give more credit. Mr. P-1 PrimeNet 1 2011-10-24 14:17
New paper from Wieb Bosma submitted to arXiv.org schickel Aliquot Sequences 4 2010-01-05 18:37
Crazy question... guido72 Software 6 2008-08-16 02:34
My Prime95 is going crazy!! xago666 Information & Answers 2 2008-03-12 18:38
Have I submitted bad data? (Or, well, about to anyway) Nazo Software 5 2005-08-06 05:44

All times are UTC. The time now is 03:54.

Sun Feb 28 03:54:20 UTC 2021 up 87 days, 5 mins, 0 users, load averages: 1.38, 1.57, 1.60

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.