Use a few PC - Page 2

VBCurtis · 2018-02-21, 00:38

By the time you've factored a dozen or two numbers with NFS and the instructions in Mr Gilchrist's beginner's guide, you might have some grasp of what each step does.
Once you have that grasp, the code in factmsieve.py becomes readable, and you learn what you need to adjust to split a large job over 2 or 3 machines. How to split the work depends on what input you're factoring, what parameters you choose, and whether both machines will be available the entire length of the project (for instance, I have on occasion used a second machine for a month to help a primary machine finish a 6-month job in less time, for something around GNFS-180 digits difficulty).

As someone else said, effort required doubles every 5-ish digits. Once you succeed with RSA-100 and RSA-110, assume the time taken scales every 10 digits just as it scaled between those two. An hour for RSA-100 is possible, 4 hrs for RSA-110. That's 16 hrs for RSA-120,..... continue as far as you wish to get an idea how long things take.

RSA-200 would then be 4^10 = 2^20 = 1,000,000 hours on a typical 4-core PC. That's an overestimate, but 20 years is a very rough estimate. Nothing that size or larger has been completed on a single PC.

So, how long would RSA-340 take, if it were possible? That's 4^14 times longer than RSA-200, or 2^28 = 250 million times longer than 20 years. This estimate aside, it's not possible with software currently available.

axn · 2018-02-21, 02:47

Quote:

Originally Posted by Geek2600

Yes,i understad.
But it's only first step,just for test how it will be work.
My target it's RSA 640.
And i have few PC with core i7.
However,now i don't understand how i must start msieve ...
If you can tell me,what exactly i must write in a argument's it will be very helpful.

See https://en.wikipedia.org/wiki/RSA_numbers

RSA-640 is a 193 digit number that has been already factored.

RSA-340 is a 340 digit number.

I understand that the naming convention is confusing.

Geek2600 · 2018-02-21, 03:20

Quote:

Originally Posted by wombatman

I would start with this: http://gilchrist.ca/jeff/factoring/n...ers_guide.html

Yes,i read it.
Sorry if my question or me a little bit stupid,but i repeat my ask for you...
What i must write as a arguments?
Just example.
Please..

wombatman · 2018-02-21, 04:01

You can run msieve -h to get all of the command line arguments. As an example, from the site I linked you to:

Code:

msieve -s example.dat -l example.log -i example.ini -nf example.fb -v -np

VBCurtis · 2018-02-21, 05:02

If you read the guide, what specific step are you confused about? What arguments are you referring to? Enter the commands as suggested in the guide, watch factorization happen. I don't know what to tell you that isn't in the guide.

If you mean "how do I split it onto two machines right now?", that's much too complicated to answer when you don't even know how to use the software packages. Do a few dozen factorizations first.

Geek2600 · 2018-02-26, 12:49

Quote:

Originally Posted by wombatman

You can run msieve -h to get all of the command line arguments. As an example, from the site I linked you to:

Code:

msieve -s example.dat -l example.log -i example.ini -nf example.fb -v -np

Thank yoy so much.
as far as I understand, to distribute the load between the cores, I have to write:

Code:

msieve -s example.dat -l example.log -i example.ini -nf example.fb -v -np 1,1000

Code:

msieve -s example.dat -l example.log -i example2.ini -nf example3.fb -v -np 1001,2000

After this command i will get 2 fb files with different results.
How i must use it?

VBCurtis · 2018-02-26, 15:24

Until you demonstrate that you have completed a full factorization by following the directions given on the beginner's page, there's not much we can tell you.

If you split the work as you indicated, you'll get 5 or 10 seconds of poly select from each machine. Why bother?

To answer your query, you would continue the factorization with whichever polynomial has the best score from among your runs.

2018-02-21, 00:38	#12
VBCurtis "Curtis" Feb 2005 Riverside, CA 4,861 Posts	By the time you've factored a dozen or two numbers with NFS and the instructions in Mr Gilchrist's beginner's guide, you might have some grasp of what each step does. Once you have that grasp, the code in factmsieve.py becomes readable, and you learn what you need to adjust to split a large job over 2 or 3 machines. How to split the work depends on what input you're factoring, what parameters you choose, and whether both machines will be available the entire length of the project (for instance, I have on occasion used a second machine for a month to help a primary machine finish a 6-month job in less time, for something around GNFS-180 digits difficulty). As someone else said, effort required doubles every 5-ish digits. Once you succeed with RSA-100 and RSA-110, assume the time taken scales every 10 digits just as it scaled between those two. An hour for RSA-100 is possible, 4 hrs for RSA-110. That's 16 hrs for RSA-120,..... continue as far as you wish to get an idea how long things take. RSA-200 would then be 4^10 = 2^20 = 1,000,000 hours on a typical 4-core PC. That's an overestimate, but 20 years is a very rough estimate. Nothing that size or larger has been completed on a single PC. So, how long would RSA-340 take, if it were possible? That's 4^14 times longer than RSA-200, or 2^28 = 250 million times longer than 20 years. This estimate aside, it's not possible with software currently available. Last fiddled with by VBCurtis on 2018-02-21 at 00:46 Reason: Fixed number errors

2018-02-21, 04:01	#15
wombatman I moo ablest echo power! May 2013 29×61 Posts	You can run msieve -h to get all of the command line arguments. As an example, from the site I linked you to: Code: msieve -s example.dat -l example.log -i example.ini -nf example.fb -v -np

2018-02-21, 05:02	#16
VBCurtis "Curtis" Feb 2005 Riverside, CA 4,861 Posts	If you read the guide, what specific step are you confused about? What arguments are you referring to? Enter the commands as suggested in the guide, watch factorization happen. I don't know what to tell you that isn't in the guide. If you mean "how do I split it onto two machines right now?", that's much too complicated to answer when you don't even know how to use the software packages. Do a few dozen factorizations first.

2018-02-26, 15:24	#18
VBCurtis "Curtis" Feb 2005 Riverside, CA 4,861 Posts	Until you demonstrate that you have completed a full factorization by following the directions given on the beginner's page, there's not much we can tell you. If you split the work as you indicated, you'll get 5 or 10 seconds of poly select from each machine. Why bother? To answer your query, you would continue the factorization with whichever polynomial has the best score from among your runs.