mersenneforum.org  

Go Back   mersenneforum.org > Factoring Projects > Factoring

Reply
 
Thread Tools
Old 2021-04-03, 00:10   #1
Unitome
 
Apr 2021

17 Posts
Default Résumé factoring?

If you keep all the log files and the poly (polynomial) file, if you exit out and start it back up again will it resume where it left off? Using the GGNFS/Msieve windows implementation.
Unitome is offline   Reply With Quote
Old 2021-04-03, 02:35   #2
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

2·2,393 Posts
Default

It depends. ggnfs/msieve are usually called by some other program- which one are you using?
If it's the python script factmsieve, and you're still in the sieving phase, it'll pick back up at the start of the Q-range you were doing when it left off (so, if you have 10,000 blocks, it'll rewind no more than 10,000).

If you're in the matrix phase, I think it restarts the matrix unless you invoke msieve manually with -ncr instead of -nc2 (-ncr = resume matrix solving, rather than restart it).
VBCurtis is offline   Reply With Quote
Old 2021-04-03, 04:02   #3
Unitome
 
Apr 2021

17 Posts
Default

Thank you! Yes factmsieve.py. Most of that went over my head as a beginner, so the answer is you don't loose all of your progress which is good. For a 140 digit number would you loose less than say 25% of your progress would you say? Assuming you are in sieving as that is the longest step afaik.

So I would do python factmsieve-0.86.py RSA140.n -ncr to resume from the log files? And does matrix refer to polynomial selection?

That said, a "pause" feature would be awesome because sometimes you need to use a processor intensive application without waiting days for the factoring to finish ;).

Last fiddled with by Unitome on 2021-04-03 at 04:09
Unitome is offline   Reply With Quote
Old 2021-04-03, 05:39   #4
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

2×2,393 Posts
Default

You should do some small jobs, like C100-110, to see how everything works before doing C140+. The matrix is the last long step.
Don't use -ncr to resume your job- just use the same command you used originally. factmsieve figures out it's partway done with the job, and picks up with the same GGNFS command it left off with.

You know how every so often the GGNFS sievers finish their work and there's some steps that show a number of relations gathered and maybe a percent done? You only lose the most recent partly-completed one of those steps.
Small jobs might only have 3-5 rounds of those GGNFS siever steps, but a C140 likely has a bunch so you're barely losing any work in the big picture- maybe 30 min on a 1-day job?
VBCurtis is offline   Reply With Quote
Old 2021-04-04, 15:03   #5
Unitome
 
Apr 2021

17 Posts
Default

Wow that's awesome! Yes, I started at RSA100, 110, 120, 130, have just completed RSA140. That took my PC about 40 hours so now am thinking how to progress from here lol, I think perhaps generating some random 140 digit numbers and benchmarking those against the 40hr RSA. With 2.5gb of log files for RSA 140, that's no joke!

Last fiddled with by Unitome on 2021-04-04 at 15:04
Unitome is offline   Reply With Quote
Old 2021-04-04, 16:20   #6
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

112628 Posts
Default

If you're asking for numbers to factor, I can keep you fed with C135-150 sized numbers from aliquot-sequence project for some amount of time (maybe a dozen numbers before I point you at the project itself and show you where to get them/ what to do other than factor one).
PM me if interested.
VBCurtis is offline   Reply With Quote
Old 2021-04-04, 20:31   #7
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

943410 Posts
Exclamation

Quote:
Originally Posted by Unitome View Post
...With 2.5gb of log files for RSA 140, that's no joke!
You might be saving some needless logs. E.g. stdout is totally expendable, you can simply run factmsieve.py ${myc140} &> /dev/null &
If you are saving stdout into a file, that is nearly definitely a waste.

On the other hand, the issue is that data files are large -- this data is indeed essential for factorization (all files with *.dat* in the name); don't delete them until your project is completed. These are not 'log files'; that is a (so to say) witness data that when it will have passed the watershed limit - will solve the factorization problem.
You can use the compressed data option - this will use approximately half the disk space, and nearly the same speed.
Batalov is offline   Reply With Quote
Old 2021-04-05, 00:27   #8
Unitome
 
Apr 2021

17 Posts
Default

Thanks very much for that, good to know what is expendable and not, and it is the dat files that are the biggest.

Running into another problem, when I make up a number to try I get the error:

Error: evaluated polynomial value polyval is not a multiple of n!

Do we have to use special numbers to work in this program or is there some setting I am missing, again RSA numbers have worked fine but numbers I come up with invariably don't work...
Unitome is offline   Reply With Quote
Old 2021-04-05, 00:42   #9
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

2×53×89 Posts
Default

Error: evaluated polynomial value polyval is not a multiple of n!
This message means that the polynomial file is inconsistent. (e.g. n: data row contradicts the c*: rows... or other data is contradictory.)
You have to start each project in separate 'namespace'. meaning: you can still use the same folder but you cannot use the same project "root"-name.

Best practice is to use separate folders. Start from a new empty folder, and start from a fresh input number. From an input number a polynomial is built first, the sieving happens, then matrix is built, then linear algebra problem is solved. You probably have data lingering from the previous project.

Quote:
Originally Posted by Unitome View Post
Do we have to use special numbers to work in this program
Any number can be factored given appropriate time.
Batalov is offline   Reply With Quote
Old 2021-04-05, 07:03   #10
Unitome
 
Apr 2021

100012 Posts
Default

Quote:
Originally Posted by Batalov View Post
Error: evaluated polynomial value polyval is not a multiple of n!
This message means that the polynomial file is inconsistent. (e.g. n: data row contradicts the c*: rows... or other data is contradictory.)
You have to start each project in separate 'namespace'. meaning: you can still use the same folder but you cannot use the same project "root"-name.

Best practice is to use separate folders. Start from a new empty folder, and start from a fresh input number. From an input number a polynomial is built first, the sieving happens, then matrix is built, then linear algebra problem is solved. You probably have data lingering from the previous project.


Any number can be factored given appropriate time.
It's very strange, I got an entire new fresh GGNFS folder (from these forums) with everything in it, put it in C: and all I did was take the RSA100.n file and changed the first digit from a 1 to a 2 and it throws the error. If I switch it back to a 1 then it works fine. So it was a brand new folder with nothing in it from a previous factorization. For some reason any RSA number works for me but any number I come up with doesn't work... Any ideas what could be going wrong?

Also I am now working on a 134 digit number by VBCurtis and it makes it through polynomial selection just fine. What is it about made up numbers that don't seem to work out?

Last fiddled with by Unitome on 2021-04-05 at 07:30
Unitome is offline   Reply With Quote
Old 2021-04-05, 08:44   #11
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
Jun 2011
Thailand

22·17·139 Posts
Default

If you took the RSA number and changed a 1 into a 2, you have a 50% chance you got a number divisible by 3, plus a ~16% chance your number is divisible by 7, etc, so a quite high chance your new number has very small prime factors. You will need to run a lot of other stuff (TF, P-1, ECM) on it to make it "NFS-ready", before attempting to find any (enough) independent relations... Try yafu on your new number and see what factors will it come with.

Last fiddled with by LaurV on 2021-04-05 at 08:46
LaurV is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
factoring 2ⁿ-2 equivalent to factoring 2ⁿ-1(I think) baih Miscellaneous Math 9 2020-09-21 07:11
OpenCL GPU P-1 Factoring and ECM Factoring xx005fs GPU Computing 3 2018-10-27 14:49

All times are UTC. The time now is 02:39.

Sun May 16 02:39:57 UTC 2021 up 37 days, 21:20, 0 users, load averages: 3.38, 3.41, 2.91

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.