mersenneforum.org  

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

Reply
 
Thread Tools
Old 2020-07-19, 17:47   #3279
Chuck
 
Chuck's Avatar
 
May 2011
Orange Park, FL

5·173 Posts
Default

Quote:
Originally Posted by kriesel View Post
Running first-time PRP without -proof, or even worse, first-time LL, by CUDALucas, prime95/mprime, Mlucas, etc is haste creating waste, by requiring more future DC, that is quite avoidable. (PRP with proof is possible on Gpuowl now, not yet released for prime95/mprime, not yet developed for Mlucas.)
Needed DC is currently a very good use of cycles, while we wait for PRP proof capability to expand.
We need to make a serious effort to get curtisc to convert to PRP.
Chuck is offline   Reply With Quote
Old 2020-07-19, 18:01   #3280
axn
 
axn's Avatar
 
Jun 2003

7·11·61 Posts
Default

Quote:
Originally Posted by Chuck View Post
We need to make a serious effort to get curtisc to convert to PRP.
1. George needs to email curtisc.
2. curtisc needs to get access to the Uni machines.
axn is offline   Reply With Quote
Old 2020-07-19, 19:00   #3281
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

37×233 Posts
Default

Quote:
Originally Posted by kriesel View Post
Then, on to 53M in October. (Looks like Fan Ming did cleanup of expired DC for 51M.)
You mean poached active assignments.
The last 51M exponent was progressing. And the 52 exponent was a few days off, but progressing. That got poached. Waiting 10 days for between milestones in no big issue. Even a month between them is nothing.

Now that the squatter is off the other exponents, the next million milestone should be achieved soon enough too.
Uncwilly is online now   Reply With Quote
Old 2020-07-19, 22:09   #3282
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

23·193 Posts
Default

Quote:
Originally Posted by Chuck View Post
We need to make a serious effort to get curtisc to convert to PRP.
Given the number of machines involved, he may want to wait until a PRP-proof-enabled version of prime95/mprime is released for general use. Update the herd once for version and preferred work, rather than two rounds.
kriesel is online now   Reply With Quote
Old 2020-07-19, 22:24   #3283
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

23×193 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
You mean poached active assignments.
I wrote what I meant. See https://mersenneforum.org/showpost.p...postcount=3274 Both the exponents listed there were seen to have been stalled since ~12 June then, and since.
kriesel is online now   Reply With Quote
Old 2020-07-20, 14:18   #3284
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

37·233 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
You mean poached active assignments.
The last 51M exponent was progressing. And the 52 exponent was a few days off, but progressing. That got poached. Waiting 10 days for between milestones in no big issue. Even a month between them is nothing.
Quote:
Originally Posted by kriesel View Post
I wrote what I meant. See https://mersenneforum.org/showpost.p...postcount=3274 Both the exponents listed there were seen to have been stalled since ~12 June then, and since.
Quote:
Originally Posted by kriesel View Post
As of now, 1 left to 51M DC milestone (estimated 3 days), and 3 more to 52M DC, one of which is checking in regularly; two seem stalled.
docustompc has been contacted for 51931577
And someone who may or may not be William P Wilkey about the other. 51791717
The last hold out for the 51M finished naturally.
One of the last 2 in the 51-52 range completed naturally.
There are no manual assignments allowed in the Cat 1 or Cat 0 range.
So Fan Ming, went outside of the PrimeNet assignment rules. If a user turns in 'manual assignment' work in a range where there are no manual assignments, they are poaching. When the assignments expired, they get sent out again. So some well meaning person that has been turning in Cat 1 and 0 assignments consistently got handed out an assignment that Fan Ming poached before the person put a single cycle to. If I recall correctly, that last assignment vaporizes once a matching DC is turned in (thus does not show up on the exponent report), I think that this happened to me once.
With the assignment rules and recycling of assignments, we should have had 52 achieved by the end of July. Again, waiting 10 or 12 days for a milestone is nothing.
Uncwilly is online now   Reply With Quote
Old 2020-07-20, 14:46   #3285
Mark Rose
 
Mark Rose's Avatar
 
"/X\(‘-‘)/X\"
Jan 2013

7×409 Posts
Default

Makes me wonder if poaching results could be held in a different table to be processed when an active assignment expires.
Mark Rose is offline   Reply With Quote
Old 2020-07-20, 17:31   #3286
ATH
Einyen
 
ATH's Avatar
 
Dec 2003
Denmark

22×3×5×72 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
If a user turns in 'manual assignment' work in a range where there are no manual assignments, they are poaching.
This may result in poaching but it is not the definition of poaching. You can also turn in manual results in Cat0/1 which had no current assignment, I do this all the time. All Cat 0/1 exponents are not assigned at once.

Poaching is turning in a result, manually or otherwise, while another person has an active assignment that has not expired.

You can poach without manual results, if you add a line to worktodo.txt you get N/A instead of assignment ID, and when Prime95/mprime finishes you will have poached the exponent.
(unless the current assignment expired before this and no new assignment was made).

Last fiddled with by ATH on 2020-07-20 at 17:36
ATH is offline   Reply With Quote
Old 2020-07-20, 18:09   #3287
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

21AD16 Posts
Default

AFAIK, the server records "Manual Assignment" for any result without an AID. The problem with doing these turning in of work between the moment it expires and when it gets assigned is that it is the wild west. People might be working behind the scenes on the same exponent. There is also cases like what happened to me recently. There was an exponent running close to expiration. I contacted the normally responsible user. The machine was not going to be able to do the work. They agreed to drop the exponent. As soon as they did, I did the self-assignment trick and got a valid AID from PrimeNet, all legit. My machine started to work on it. Some person came in after the exponent would have expired and turned in the DC (done on a GPU). So, I lost several days of time. Letting the system work saves cycles.

Not everyone who poaches cares to be the most useful to the project. I think some are not on the forum.

If we tried coordinate who would spearfish the tail-end DC's that might help. Or have a user or 5 that get all of the expired Cat 0 assigned to them the instant that they expire. The server could email them the their assignment and any known interim residues.
Uncwilly is online now   Reply With Quote
Old 2020-07-20, 18:28   #3288
ATH
Einyen
 
ATH's Avatar
 
Dec 2003
Denmark

22·3·5·72 Posts
Default

I try to grab assignments just when they expire so I get an AID like these, and then it is no longer poaching. If I turn in results without AID, I make sure there is no active assignments to someone else.

Exceptions recently was the endless "patgie" assignments for 6+/12+ months without progress and the unneeded TF assignments in the DC range. Both of which most of us agreed was "borderline ok" to poach.

Last fiddled with by ATH on 2020-07-20 at 18:56
ATH is offline   Reply With Quote
Old 2020-07-21, 04:01   #3289
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
Jun 2011
Thailand

2×3×31×47 Posts
Default

Quote:
Originally Posted by Mark Rose View Post
Makes me wonder if poaching results could be held in a different table to be processed when an active assignment expires.
It wouldn't be beneficial. Poacher's mind is twisted (don't ask me how I know! ), if he doesn't see his result displayed, he will assume something went wrong (or just forget the history, and spot the same candidate again) and he will try again and again, wasting even more resources, when he would be better "contributing" to different "assignments". So, let them have at it, display the results, but don't credit it, and don't display the name (or say "poacher" with lower case, hihi). That should deter most of them.

On the other hand, the summary of the last posts, is that in Ken's world, the poaching is bad only when is done by me

Last fiddled with by LaurV on 2020-07-21 at 04:03
LaurV 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 17:51.

Sun Sep 27 17:51:29 UTC 2020 up 17 days, 15:02, 0 users, load averages: 1.43, 1.46, 1.50

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.