mersenneforum.org  

Go Back   mersenneforum.org > New To GIMPS? Start Here! > Information & Answers

Reply
 
Thread Tools
Old 2021-05-08, 10:41   #1
Dalybe
 
May 2021

7 Posts
Default Complete GIMPS Beginner

Hi Everybody,

I've decided to join the quest for discovering Mersenne primes and incidentally - this is my foray into posting on a forum (so please bare with me!)

This has probably been covered many times over already - but I've had a look through the forum and haven't had much luck trying to find an answer to my questions.
If someone could help me out please that would be much appreciated!

OK, so firstly - is there documentation / a forum thread that I can read that gives me a detailed explanation of what I am looking at in terms of my personal assignment details?

For example, I see columns like 'CPU name' , 'Exponent' and 'Work Type' but I am unclear as to what they mean (especially because the number being displayed in 'Exponent' column is different from the one that I am testing in Prime95?).

This is all for now - but I'm sure I'll have more questions ! Thanks in advance for any advice.

Dalybe
Dalybe is offline   Reply With Quote
Old 2021-05-08, 11:06   #2
moebius
 
moebius's Avatar
 
"CharlesgubOB"
Jul 2009
Germany

27216 Posts
Default

use https://www.mersenne.org/workload/
to see the progress off all your assignments
you can update the progress in prime95 manually with Advanced>manual communication>Send new expected completion date

Last fiddled with by moebius on 2021-05-08 at 11:09
moebius is offline   Reply With Quote
Old 2021-05-08, 12:28   #3
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

10111000111102 Posts
Default

Welcome to the forum. There are some pretty good documentation files shipped in the zip file with prime95 or mprime. Read them, try some things, use the reference info found at https://mersenneforum.org/showthread...922#post521922, and have fun!
kriesel is offline   Reply With Quote
Old 2021-05-08, 12:51   #4
Dalybe
 
May 2021

7 Posts
Default

That's great - thank you @kriesel and @moebius . I will do a some more reading and investigating, thanks!
Dalybe is offline   Reply With Quote
Old 2021-05-08, 18:57   #5
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

2×3×5×337 Posts
Default

One of the things to note:
When doing certain types of tests, the test hast to run to the end or a major point before any results are known.

PRP tests (the current standard way to check a number for being prime) is like this
LL tests (which is what has been what we used to use all the time in the past) is like this
P-1 tests (which look for factors for numbers, so we don't have to do the longer tests) have a stage 1 and stage 2. The test can find a factor at the end of either, but not in the middle.
TF (trial factoring) can find a factor at any time (essentially), thus can stop whena factor is found.
Uncwilly is offline   Reply With Quote
Old 2021-05-10, 15:06   #6
LOBES
 
Mar 2019
USA

3·23 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
One of the things to note:
When doing certain types of tests, the test hast to run to the end or a major point before any results are known.

PRP tests (the current standard way to check a number for being prime) is like this
LL tests (which is what has been what we used to use all the time in the past) is like this
P-1 tests (which look for factors for numbers, so we don't have to do the longer tests) have a stage 1 and stage 2. The test can find a factor at the end of either, but not in the middle.
TF (trial factoring) can find a factor at any time (essentially), thus can stop whena factor is found.
I'm sure the answer to these questions is found in the various detailed descriptions of the various methods....but frankly after the first couple of sentences in those my brain turns off. I've always been curious about the following, and since this is a newbie thread, seemed appropriate:

1. Is the length of time a P-1 takes to complete lowered at all by how high the TF has been done? It always says" Assuming no factors below 2^76" (for example). If that same exponent was instead, say, factored to 2^78 before running the P-1, would the P-1 test take less time?

2. Am I correct that if a PRP test is performed on an exponent that does not have a P-1 test done yet, that it will do the P-1 prior to running the PRP? So, running a PRP on an exponent that already has a P-1 would take less time since it doesn't have to run that test first?

3. In the most simplest terms possible, can someone explain why a TF on lower exponents take LONGER than on much much larger exponents? My GTX 1080 Ti burns though high level exponents, but takes significantly longer for lower ones.
LOBES is offline   Reply With Quote
Old 2021-05-10, 15:44   #7
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

2·11·269 Posts
Default

Quote:
Originally Posted by LOBES View Post
1. Is the length of time a P-1 takes to complete lowered at all by how high the TF has been done? It always says" Assuming no factors below 2^76" (for example). If that same exponent was instead, say, factored to 2^78 before running the P-1, would the P-1 test take less time?
Yes on prime95 / mprime, which determines for itself what P-1 bounds to use to optimize overall project progress rate based on exponent, indicated prior TF done in the worktodo line, Dickman's function, etc. No on gpuowl in my experience, which uses a default set of bounds for a wide range of exponents. (That may have changed in gpuowl v7.x.) But please use the optimal TF levels and P-1 bounds.
Quote:
2. Am I correct that if a PRP test is performed on an exponent that does not have a P-1 test done yet, that it will do the P-1 prior to running the PRP? So, running a PRP on an exponent that already has a P-1 would take less time since it doesn't have to run that test first?
Correct on prime95 / mprime, and recent versions of Gpuowl, if the worktodo line reflects that. But the effect is minor; ~1/35 of PRP run time on a Radeon VII, for example. One caveat; it's my understanding, and IIRC George confirmed it a while back, a P-1 run to inadequate bounds is treated the same as no previous P-1 when the PRP work assignment is issued. So please use adequate bounds. Stage-1-only is generally inadequate. Enable adequate ram usage in prime95/mprime so the second stage P-1 will run with adequate B1 & B2 for the exponent. (See prime95's readme for some guidance.)
Quote:
3. In the most simplest terms possible, can someone explain why a TF on lower exponents take LONGER than on much much larger exponents? My GTX 1080 Ti burns though high level exponents, but takes significantly longer for lower ones.
For the same bit level, say going from 73 to 74 bits, since factors are of form f = 2 k p + 1, 273< 2 k p + 1 < 274, where p is the Mersenne number prime exponent, for large p there are fewer k values to test, while for small p there are many more k values to test. A 10M exponent has about a hundred times more potential factors to try than a 1G exponent for the same bit level.

But large exponents take longer to trial factor overall, since their optimal highest bit level to TF is higher than for smaller exponents. TF effort for a bit level is about proportional to 2level/exponent. There are several other details that go into it; density of primes declining further up the number line, GPU factoring kernels' complexity increasing with bit level, etc. Other readers that want more in-depth info can read this.
For tables, charts, scaling fits, the pdf attachment here contains those.

Last fiddled with by kriesel on 2021-05-10 at 15:59
kriesel is offline   Reply With Quote
Old 2021-05-10, 15:48   #8
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
"name field"
Jun 2011
Thailand

100110010100002 Posts
Default

1. No. For a specific B1/B2 pair, P-1 will take the same amount of time, with or without any TF done. However its chances to find a factor are different. Prime95 and gpuOwl will calculate different B1/B2 knowing the TF level, in such a way to maximize your chances to find a factor.

2. It's complicate. Your PRP alone will take the same amount of time regardless of the fact that P-1 was done or not. Also viceversa, the P-1 will take the same amount of time alone, regardless the fact that PRP was done or not. However, where it gets complicate is the fact that P-1 may found a factor, therefore saving the long PRP run, and on the long run, your total time spent doing PRP could be lower. Also, new "toys" like gpuOwl, etc, combine the PRP and P-1 in one (because in the beginning the iterations are common) so the P-1 may come "for free" if no factor is found.

3. Factors of 2^p-1 are of the form 2kp+1. So, factors of 2^11-1 can be apart a multiple of 22. Factors of 2^67-1 can be a multiple of 67+67=134 apart. Factors of 2^101-1 can be 202 apart. If you TF the two exponents (67 and 101) from, say, 2^20 to 2^21, how many possible candidates you have in each case?

Last fiddled with by LaurV on 2021-05-10 at 15:51
LaurV is offline   Reply With Quote
Old 2021-05-10, 21:03   #9
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

2·11·349 Posts
Default

Quote:
Originally Posted by LOBES View Post
1. Is the length of time a P-1 takes to complete lowered at all by how high the TF has been done? It always says" Assuming no factors below 2^76" (for example). If that same exponent was instead, say, factored to 2^78 before running the P-1, would the P-1 test take less time?.
Since you two different answers on this question, the answer is yes. More TF means prime95 will choose smaller B1/B2 bounds which will take less time. If for some reason someone did an inordinate amount of TF - say to 2^89 - prime95 may skip P-1 altogether.
Prime95 is offline   Reply With Quote
Old 2021-05-11, 23:33   #10
MattcAnderson
 
MattcAnderson's Avatar
 
"Matthew Anderson"
Dec 2010
Oregon, USA

3B916 Posts
Default

Welcome to the team, Dalybe !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
MattcAnderson is offline   Reply With Quote
Old 2021-05-14, 15:31   #11
Dalybe
 
May 2021

7 Posts
Default

Hello again everybody - and thanks for making me feel very welcome!

Ok, so I have completed (at least I think I have completed) my first assignment.
It turns out that my first one wasn't prime (surprise surprise!) and now I have been given a new one to check.
What I am confused about however is why on my assignments tab it is still displaying information from the initial one which has completed? Is there another series of checks or something that has to happen before this is updated?

Additionally, the 'Last Update' column says 2021-05-08 - but apparently the check was complete on 2021-05-13. Any help or advice would be much appreciated! Cheers!

Dalybe
Dalybe is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Beginner prime search marcusvdl And now for something completely different 9 2018-05-04 01:26
A beginner need help!!! VinhAn Information & Answers 3 2007-08-09 19:57
f14 complete masser Sierpinski/Riesel Base 5 2 2006-04-23 16:05
Good Linux distro for a beginner ThomRuley Linux 17 2004-08-30 01:12
61.5 thru 62m complete to 2^60 nitro Lone Mersenne Hunters 0 2003-12-07 13:50

All times are UTC. The time now is 08:47.


Wed Dec 1 08:47:55 UTC 2021 up 131 days, 3:16, 1 user, load averages: 1.36, 1.36, 1.31

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.