mersenneforum.org Is the website outdated, or me stupid?
 User Name Remember Me? Password
 Register FAQ Search Today's Posts Mark Forums Read

 2021-04-16, 14:02 #1 drkirkby   "David Kirkby" Jan 2021 Althorne, Essex, UK 26×7 Posts Is the website outdated, or me stupid? First, I'm not a mathematician, so its probably the latter. According to this page on the maths https://www.mersenne.org/various/math.php when discussing trial factoring, it says "Now the only question remaining is how much trial factoring should be done? The answer depends on three variables: the cost of factoring, the chance of finding a factor, and the cost of a primality test. The formula used is: factoring_cost < chance_of_finding_factor * 2 * primality_test_cost That is, the time spent factoring must be less than the expected time saved." Where does the factor of 2 come from? Is it outdated now PRP tests are used, so that a LL test does not have to be performed twice? Or is it still current? I can't seem to get my head around that one - part of me thinks it is right, but part of me thinks it is wrong. Being the website admin, director, loo cleaner etc for my own company, https://www.kirkbymicrowave.co.uk/ I know what a nightmare it is trying to keep a website up to date, but I'm not even convinced it is outdated, but suspect it might be. Dave Last fiddled with by drkirkby on 2021-04-16 at 14:03
2021-04-16, 14:11   #2
Uncwilly
6809 > 6502

"""""""""""""""""""
Aug 2003
101×103 Posts

13·19·41 Posts

Quote:
 Originally Posted by drkirkby Where does the factor of 2 come from? Is it outdated now PRP tests are used, so that a LL test does not have to be performed twice?
It is outdated. The 2 is from 2 L-L tests. Mersenne.ca and GPU72 have started to factor in the savings, wrt bit level. Since TFing each bit level is double the effort of the bit level below, that translates to do 1 less bit level.

2021-04-16, 14:12   #3
axn

Jun 2003

22·1,301 Posts

Quote:
 Originally Posted by drkirkby Where does the factor of 2 come from? Is it outdated now PRP tests are used, so that a LL test does not have to be performed twice? Or is it still current? I can't seem to get my head around that one - part of me thinks it is right, but part of me thinks it is wrong.
Yes, the factor of 2 is based on the 2 LL tests needed to conclusively prove an exponent is composite. With PRP+CERT, the factor of 2 should become 1.03 (or something similar).

2021-04-16, 20:11   #4
kriesel

"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

17×349 Posts

Quote:
 Originally Posted by axn Yes, the factor of 2 is based on the 2 LL tests needed to conclusively prove an exponent is composite. With PRP+CERT, the factor of 2 should become 1.03 (or something similar).
Really more like for LL, 2(1+e+e2+...] where e=error rate per primality test is ~0.01 for ~p=100M with Jacobi check, double~0.02 without Jacobi check, and e is a strong function of exponent, or of runtime which is roughly 2.1 power of exponent. So 2.02 for LL with Jacobi check, 2.04 without, for 100M exponent, but ~2.5 for 100Mdigit, and ~19. for gigabit, and enormous for gigadigit or higher, as independent runs, due to increasingly large probability of error. (Background to support the LL error rate estimates is available in attachments at https://www.mersenneforum.org/showpo...6&postcount=14)

For PRP with GEC and proof, the numbers are much better; ~1.01, including GEC which is ~0.2% for block size 1000, thanks to the great reduction in verification by the VDF based proof generation and CERT, and the quick error detection and recovery along the way of GEC.
The value depends on the proof power and whether the initial verification succeeds.
For the usual case, proof power 8 or 9, and successful verification the first time, total test and verification effort is <1.01 times a single PRP test without proof.

Attached Files
 prp preferable.pdf (21.8 KB, 51 views) proof cost.pdf (21.4 KB, 55 views)

 2021-04-17, 00:07 #5 drkirkby   "David Kirkby" Jan 2021 Althorne, Essex, UK 1110000002 Posts Who could update the website?
2021-04-17, 23:56   #6
LOBES

Mar 2019
USA

6910 Posts

Quote:
 Originally Posted by drkirkby Who could update the website?
If whomever does update the site, can you please add one more column to the following page:

https://www.mersenne.org/worktypes/

That references the numerical value of the work preference in prime95?

2021-04-18, 02:05   #7
Uncwilly
6809 > 6502

"""""""""""""""""""
Aug 2003
101×103 Posts

236178 Posts

Quote:
 Originally Posted by LOBES That references the numerical value of the work preference in mprime95?
Fixed that for you.

2021-04-18, 08:58   #8
drkirkby

"David Kirkby"
Jan 2021
Althorne, Essex, UK

26×7 Posts

Quote:
 Originally Posted by LOBES If whomever does update the site, can you please add one more column to the following page: https://www.mersenne.org/worktypes/ That references the numerical value of the work preference in prime95?
Do you mean like like the following?

Code:
Use the following values to select a work type:
0 - Whatever makes the most sense
150 - First time prime tests
152 - World record sized numbers to prime test
151 - Double-check prime tests
2 - Trial factoring
4 - P-1 factoring
153 - 100 million digit numbers to prime test
160 - First time PRP on Mersenne cofactors
161 - Double-check PRP on Mersenne cofactors
5 - ECM for first factors of Mersenne numbers
8 - ECM on Mersenne cofactors
6 - ECM on Fermat numbers
1 - Trial factoring to low limits
Uncwilly says he has fixed the page for you, but if you mean what I think you mean, the page is not fixed in the way you want.

2021-04-18, 10:10   #9
S485122

"Jacob"
Sep 2006
Brussels, Belgium

33348 Posts

Quote:
 Originally Posted by Uncwilly Fixed that for you.
The numerical values are used in the prime.txt configuration file for both mprime and Prime95 ... and to communicate with PrimeNet. But only the mprime user interface confronts one with them directly.

It would indeed be a good idea to add that column to the table on the PrimeNet Assignment Work Types page, the v5.0 PrimeNet Web API 0.97 page can also benefit by being updated ;-)

Jacob

2021-04-18, 14:17   #10
Uncwilly
6809 > 6502

"""""""""""""""""""
Aug 2003
101×103 Posts

13×19×41 Posts

Quote:
 Originally Posted by drkirkby ]Uncwilly[/URL] says he has fixed the page for you, but if you mean what I think you mean, the page is not fixed in the way you want.
I "fixed" the mprime vs Prime95 reference. With mprime you need to use the numbers, Prime95 not so much. I did not fix the website. Only George, Aaron, and James (a little) have access to make changes to Mersenne.org.

2021-05-28, 03:24   #11
Serpentine Vermin Jar

Jul 2014

31×107 Posts

Quote:
 Originally Posted by S485122 The numerical values are used in the prime.txt configuration file for both mprime and Prime95 ... and to communicate with PrimeNet. But only the mprime user interface confronts one with them directly. It would indeed be a good idea to add that column to the table on the PrimeNet Assignment Work Types page, the v5.0 PrimeNet Web API 0.97 page can also benefit by being updated ;-) Jacob
Sorry, I haven't been keeping up with forum posts lately and just saw this.

I've updated the work types page to include the numerical value. I didn't realize that info was used or useful to anyone.

As for the API page, it's pretty confusing no matter how you slice it. My "moon shot" goal would be to replace the API with something new, but that's heavily dependent on the time involved and coordinating with George, since it involves client updates as well.

 Similar Threads Thread Thread Starter Forum Replies Last Post Dylan14 Msieve 2 2020-09-27 18:49 petrw1 Hardware 11 2013-01-16 02:45 firejuggler Lounge 9 2011-02-20 22:07 Biggles Prime Sierpinski Project 3 2006-02-07 22:50 fropones Math 2 2003-05-28 00:44

All times are UTC. The time now is 04:55.

Wed Dec 8 04:55:04 UTC 2021 up 137 days, 23:24, 1 user, load averages: 1.78, 1.57, 1.58