mersenneforum.org  

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

Reply
 
Thread Tools
Old 2019-08-11, 18:27   #1
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

D5616 Posts
Default Duplication of TF, P-1, etc. (please don't)

Please apply computing resources efficiently, avoiding needless duplication.

Unless you know the computing hardware used to do the initial TF was defective, please do not rerun the same exponent/bitlevels that have already been run. There are many examples of duplication of same tf levels same exponent by different users.

This is how it should look for more recent exponents; once for each bitlevel range.
https://www.mersenne.ca/exponent/85575233
But if the lower bit levels drop off the list, that does not mean they were not performed. They were, and then were subsequently dropped from the database years later.

An example of wasteful duplication is
https://www.mersenne.org/report_expo...exp_hi=&full=1

An example of an exponent that had full TF but subsequently some being removed from the database is 38000009

It also seems wasteful to do additional TF and P-1 after an exponent has matching composite primality test results, as in http://www.mersenne.ca/exponent/48000467
https://www.mersenne.org/report_expo...exp_hi=&full=1

In some examples, the <64 bits was not skipped, it's just they're 8 or more years old and have been removed from the database for size reasons.

This exponent had duplicated TF, duplicated P-1 to the same bounds, and triple matching LL tests, so wasted efforts in all 3 computation types. https://www.mersenne.org/report_expo...exp_hi=&full=1

Two matching LL or 2 matching PRP are intended. Three or more matching are generally considered a waste, except for the initial confirmation of a newly discovered Mersenne prime.

This exponent has an extraordinary amount of factoring and LL testing effort after May 2007 when matching LL tests were completed. Duplication of past tf, TF to much higher bit levels, P-1 factoring, and 4 unneeded LL tests, including two by the same person.
https://www.mersenne.org/report_expo...exp_hi=&full=1

Unnecessary unproductive duplication of effort slows the progress of GIMPS toward finding the next Mersenne prime.

Last fiddled with by kriesel on 2019-08-11 at 18:28
kriesel is offline   Reply With Quote
Old 2019-08-11, 23:23   #2
ixfd64
Bemusing Prompter
 
ixfd64's Avatar
 
"Danny"
Dec 2002
California

2·32·53 Posts
Default

There's an unofficial sub-project to get less than 20 million exponents without known factors. This is the main reason some of us are doing TF and P-1 on known composite numbers.

Also, a triple check often isn't intentional. If an assignment expires and the exponent is assigned to someone else, and then the original assignment is completed anyway, then this could result in more than two LL results.

Last fiddled with by ixfd64 on 2019-08-11 at 23:25
ixfd64 is offline   Reply With Quote
Old 2019-08-12, 01:22   #3
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

22×32×5×72 Posts
Default

Quote:
Originally Posted by kriesel View Post
It also seems wasteful to do additional TF and P-1 after an exponent has matching composite primality test results, as in http://www.mersenne.ca/exponent/48000467 (https://www.mersenne.org/report_expo...exp_hi=&full=1)
It depends on what one's goals are...

Some like finding factors. And for the example you provided, the first P-1 run only did Stage 1, while the second run did an appropriately deep P-1 with both Stages 1 and 2.

Quote:
Originally Posted by kriesel View Post
Unnecessary unproductive duplication of effort slows the progress of GIMPS toward finding the next Mersenne prime.
Again, not all of us here are after finding the next MP, although *many* put in a ***considerable*** amount of effort helping others do so.
chalsall is online now   Reply With Quote
Old 2019-08-12, 02:11   #4
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

17·251 Posts
Default

Quote:
Originally Posted by ixfd64 View Post
There's an unofficial sub-project to get less than 20 million exponents without known factors. This is the main reason some of us are doing TF and P-1 on known composite numbers.
Yes, that would be my brain child. And I am pleased with the support it has received.

While none of this factoring will find a prime; much of the work done here has eliminated DC work.
petrw1 is offline   Reply With Quote
Old 2019-08-12, 02:32   #5
masser
 
masser's Avatar
 
Jul 2003

33×47 Posts
Default

Some of the excessive duplication could be the result of people testing their hardware and/or software. The database is chock full of results that can be used to verify a user's setup.
masser is offline   Reply With Quote
Old 2019-08-12, 02:46   #6
ixfd64
Bemusing Prompter
 
ixfd64's Avatar
 
"Danny"
Dec 2002
California

225010 Posts
Default

Quote:
Originally Posted by masser View Post
Some of the excessive duplication could be the result of people testing their hardware and/or software. The database is chock full of results that can be used to verify a user's setup.
M1000003 seems to be a popular choice for this purpose.

Last fiddled with by ixfd64 on 2019-08-12 at 02:46
ixfd64 is offline   Reply With Quote
Old 2019-08-12, 04:36   #7
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

2×3×569 Posts
Default

mersenneforum.org > Great Internet Mersenne Prime Search
(not Great Internet Mersenne Factor Search)
Factors are a means to an end. The end is primes discovery.
That's the primary goal.

Repeating small-fft tests of years ago is not as good a test as the built in tests of the actual fft sizes used for the exponents to be run in the future.

Last fiddled with by kriesel on 2019-08-12 at 05:07
kriesel is offline   Reply With Quote
Old 2019-08-12, 05:11   #8
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

23×3×17×19 Posts
Default

My kit, my choice.

If I want to burn my electrons on doing "useless" work I can. I can also join a different project.

Yes, at times people may not understand that they are wasting FLOPs. But sometimes they do and choose to do it anyway. Your 'hail mary' factoring runs on samuel's numbers are like that. Anytime that someone claims to have found the next prime, people fire up redundant TF and P-1 efforts, for the same reason you did your work. And it is frequent that people start from the lowest bit levels.
Uncwilly is offline   Reply With Quote
Old 2019-08-12, 05:49   #9
ATH
Einyen
 
ATH's Avatar
 
Dec 2003
Denmark

2×13×107 Posts
Default

Quote:
Originally Posted by kriesel View Post
In some examples, the <64 bits was not skipped, it's just they're 8 or more years old and have been removed from the database for size reasons.
They are not removed, it is a bug they are not showing up. I tested a small range of exponents just to check if any factors were hidden there.

I might test more in the future but no plans right now. As Uncwilly sajd it is everyone's right to "waste" own their resources if they want, and it might is probably not wasteful in their eyes.

Last fiddled with by ATH on 2019-08-12 at 05:51
ATH is offline   Reply With Quote
Old 2019-08-12, 06:58   #10
GP2
 
GP2's Avatar
 
Sep 2003

A1216 Posts
Default

Quote:
Originally Posted by ATH View Post
Quote:
Originally Posted by kriesel View Post
In some examples, the <64 bits was not skipped, it's just they're 8 or more years old and have been removed from the database for size reasons.
They are not removed, it is a bug they are not showing up. I tested a small range of exponents just to check if any factors were hidden there.
Note that TJAOI is currently finding factors of bit length 65.97 and should finish up to 66 very soon. Therefore there is not much point doing TF to less than 67 bits.
GP2 is offline   Reply With Quote
Old 2019-08-12, 07:30   #11
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

2·3·569 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
My kit, my choice.

If I want to burn my electrons on doing "useless" work I can. I can also join a different project.

Yes, at times people may not understand that they are wasting FLOPs. But sometimes they do and choose to do it anyway. Your 'hail mary' factoring runs on samuel's numbers are like that. Anytime that someone claims to have found the next prime, people fire up redundant TF and P-1 efforts, for the same reason you did your work. And it is frequent that people start from the lowest bit levels.
Sure, everyone's a volunteer.

It's not about making anyone wrong. It's about having the awareness to choose well.

Nobody is making exceptional claims about 1000003. The twentieth or hundredth duplicated test on it does not advance the project. Repeated no-factor P-1 factorings of the same exponent to the same bounds by the same user are not good tests even of different hardware and software.There's a list of known exponents, bounds, and factors, for that. And no utility or benefit to reporting duplicate factoring runs to the primenet server and cluttering it with duplicates.

But exponents like those of samuel, cochet and other claimants would get tested eventually anyway. They just got addressed earlier than they otherwise would. The samuel's exponents runs I made also were experience with unusually high tests_saved values in prime95 P-1. I'm careful to make many of my runs ahead of the pack do double or triple duty or more. For example, P-1 runs that determine the limits of CUDAPm1 on a specific gpu model, that probes the software's capability, that establishes run time scaling, generating bug reports; documenting limits and workarounds, and preparing the way for other scouting runs (PRP or LL software) ahead of the pack also. Runs are staggered and well spaced, the opposite of duplicated effort. And similarly testing gpuOwL and prime95 releases in various ways, and providing feedback to the authors. As do others.

We've had a variety including some pretty extreme cases of wasted cycles, due to lack of awareness of new users. (A cpu year wasted on primality testing a 100Mdigit exponent that had already been factored is one example.) This thread could help that awareness, and help efficiency. Including for some secondary goals.

The learning curve is long. Over 150 separate reference posts in https://www.mersenneforum.org/showthread.php?t=24607 and growing.

Last fiddled with by kriesel on 2019-08-12 at 07:30
kriesel is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Duplication of Effort for Smaller Aliquot Sequences EdH Aliquot Sequences 3 2018-04-17 13:31
Duplication of work: local vs. db EdH Aliquot Sequences 2 2010-12-31 04:30
2801^79-1; thoughts on duplication sampling fivemack Factoring 0 2010-04-15 22:23

All times are UTC. The time now is 22:11.

Sun Mar 29 22:11:23 UTC 2020 up 4 days, 19:44, 2 users, load averages: 1.23, 1.22, 1.28

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.