mersenneforum.org  

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

Reply
 
Thread Tools
Old 2020-06-22, 14:25   #3257
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

859410 Posts
Default

Quote:
Originally Posted by Viliam Furik View Post
I would say, that if assignment is TF and blocking range, simply do the test manually without assignment, no? What's the point in waiting for it...
This is a problem that is best dealt with at an admin level. It is better to let the project managers manage the project. Faults in the system should be addressed, not worked around.

Quote:
Originally Posted by KEP View Post
Point is many, one very obvious one, could be that people slams the door and either leave entirely and afterwards contribute 0 days worth of work - another "less impacting" consequence could be, that users in order to avoid wasting their available GPU ressources, moves far far far far ahead of the wavefront and therefor doesn't contribute in any short term.
...
The overall goal is n=1G (most likely beyond that), so we have plenty of work available and really no need to poach. Any milestone will and should rightfully be completed in due time, without poaching.
All users should work in a manner that does not disrupt the flow of the project. If a user sings up and selects their favorite work type and lets PrimeNet hand out assignments as it sees fit, then the project managers can manage the project. If the user signs up with GPU72 and takes the assignments that are handed out, the work flow there can be managed.

When too many people get too desirous to fulfill there specific desires (like patige and horde certain exponents or others that want to work for the credits only without practically helping with the projects current needs), then the project does not work as smoothly as it should. Those of us that have signed up for and just chew through the lowest DC's are doing the unglamourous work that keeps the project moving. Will we find a prime, not likely. Will we get our cycles wasted by poachers, sometimes (way too often). But, each phase needs to be done in turn. The assignment rules help with that. Each phase should be done and cleared before the next step arrives. Care is being taken to work on that by those that are trying to manage the project. As new resources and opportunities present themselves, they are trying to put them to the best use to move the project forward as fast as possible.

We live and learn. It turns out that machines that were doing LMH-H tasks like low level factoring of exponents above 150,000, would have helped more by doing DC's, PRP-CF's, and P-1 (on some machines).

TF is supposed to save LL/PRP effort. It should never hold it up (either front end of routine PrimeNet assignments or tail-end). As more and different resources have become available, it made sense to do extra factoring ahead of the DC's in some cases. Again, that should not hold up the DC's rather accelerate them. And the TF's should work hand in glove with P-1. Sometimes it has made sense to do P-1 in the middle of the late stages of TF, sometimes it makes sense to do it after. Again, this is high level project management.
Uncwilly is offline   Reply With Quote
Old 2020-06-22, 16:20   #3258
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

3·13·113 Posts
Default

Uncwilly, mostly a bullseye response there. A couple of quibbles;
I think you mean 150M not 150,000 there.
There will be some rare exceptions to simply working at the wavefronts. There is utility in having some scouting ahead of wavefronts occur, so such things as fft size boundaries get tested a bit more than the programmers have time and inclination for, and excessive errors from being a little too aggressive with some fft size get found and fixed before the wavefront reaches there.
Poaching and hoarding are both inefficient and discouraged. Poaching DC is even more wasteful than poaching first tests. We also see poaching of TF on ocasion which is also wasteful, though usually in smaller units per assignment.
While the minor milestones may be fun to check off, and by indicating steady progress during a possibly long period between discoveries of primes, may help morale, they're not very important themselves, and are pretty arbitrary. (Consider milestones expressed in base 16 or 8.)

Last fiddled with by kriesel on 2020-06-22 at 16:26
kriesel is offline   Reply With Quote
Old 2020-06-22, 16:57   #3259
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

219216 Posts
Default

Quote:
Originally Posted by kriesel View Post
I think you mean 150M not 150,000 there.
Yep. All of those low and slow cycles could have had us a few million further up on the DC's or many more fully factored exponents found (through PRP-CF).
Quote:
There will be some rare exceptions to simply working at the wavefronts.
Yes, there is always important development work.
Quote:
While the minor milestones may be fun to check off, and by indicating steady progress during a possibly long period between discoveries of primes, may help morale,
But, the morale issue and seeing that a project is progressing can keep people working rather searching for a shinier toy. Also, helping out ticking off milestones can cause participants to maximize their efforts (more cycles).

Last fiddled with by Uncwilly on 2020-06-22 at 16:59
Uncwilly is offline   Reply With Quote
Old 2020-06-22, 17:18   #3260
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

3×13×113 Posts
Default LL/PRP mix

https://www.mersenne.org/assignments...xtf=1&exdchk=1 currently lists 408 assignments toward the 93M first-primality test milestone. A surprisingly low fraction, 15/408 = 3.7% are PRP assignments, the other 96.3% are LL

Last fiddled with by kriesel on 2020-06-22 at 17:19
kriesel is offline   Reply With Quote
Old 2020-06-23, 15:12   #3261
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

440710 Posts
Default 3K to 96M

If I read the Work Distribution Map correctly, there are only 3019 first tests to go to 96M.
https://www.mersenne.org/primenet/
1139 to 92M
237 more to 93M
1091 more to 94M
325 more to 95M
227 more to 96M

Last fiddled with by kriesel on 2020-06-23 at 15:12
kriesel is offline   Reply With Quote
Old 2020-06-24, 15:29   #3262
ryanp
 
ryanp's Avatar
 
Jun 2012
Boulder, CO

181 Posts
Default

Quote:
Originally Posted by kriesel View Post
22 to go to 51M.
The 52M will take longer; Ryan Propper has 12 with no progress and 44 days to expiration. 53M this year looks probable.
My doublechecks should be submitted now.
ryanp is offline   Reply With Quote
Old 2020-06-25, 00:32   #3263
dcheuk
 
dcheuk's Avatar
 
Jan 2019
Pittsburgh, PA

3·7·11 Posts
Default

Quote:
Originally Posted by ryanp View Post
My doublechecks should be submitted now.
Sorry about 51168703 or perhaps other ones too. Was just trying to help hitting the milestone.

My bad.

Last fiddled with by dcheuk on 2020-06-25 at 00:32
dcheuk is offline   Reply With Quote
Old 2020-06-25, 01:15   #3264
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

100001100100102 Posts
Default

Quote:
Originally Posted by dcheuk View Post
Sorry about 51168703 or perhaps other ones too. Was just trying to help hitting the milestone.

My bad.
Yes, your bad. Wait for PrimeNet to hand out assignments.
Instead of wasting cycles poaching. Spend a couple of minutes on PrimeNet looking at who has fallen behind or might hold things up. Then look on the forum and try to match a user name to the PrimeNet ID. Then PM the person. I have done that several times usually with good results (like with Ryan).
Uncwilly is offline   Reply With Quote
Old 2020-06-25, 03:57   #3265
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

3·1,459 Posts
Default

Quote:
Originally Posted by ryanp View Post
My doublechecks should be submitted now.
We miss you ... and your voluminous output.
petrw1 is offline   Reply With Quote
Old 2020-06-25, 21:35   #3266
lycorn
 
lycorn's Avatar
 
Sep 2002
Oeiras, Portugal

1,399 Posts
Default

Quote:
Originally Posted by petrw1 View Post
We miss you ... and your voluminous output.
Seconded.
lycorn is offline   Reply With Quote
Old 2020-06-27, 10:28   #3267
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

131358 Posts
Default

Why are there still 10 TF assignments for RobertKazan?
https://www.mersenne.org/assignments...xp_hi=52000000

No one has yet been assigned a DC for those 10.
retina is online now   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 13:18.

Sat Sep 19 13:18:02 UTC 2020 up 9 days, 10:29, 1 user, load averages: 1.85, 1.58, 1.51

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.