mersenneforum.org  

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

Reply
 
Thread Tools
Old 2003-09-29, 14:54   #12
GP2
 
GP2's Avatar
 
Sep 2003

22×3×5×43 Posts
Default

Quote:
Originally posted by Complex33
Another suggestion for those interested in and who have the means to P-1 test exponents would be to use PrimeNet to request a block of DC's and turn the sequentalwork switch on in prime.ini.
That's a very good point.

Let me try to formally spell it out step-by-step along with a way to semi-automate it:

- Stop Prime95 (or mprime)
- Edit prime.ini to add:
Time=0:00-1:00,9:00-24:00
just after the UserID line (but pick slightly different times).
- Restart Prime95

- Copy the entire Prime95 directory to a second directory.
- In the second directory, clear out worktodo.ini
- In the second directory, edit prime.ini as follows:
Add Time=1:00-9:00 just after the UserID line (but pick slightly different times).
Set WorkPreference=4 (for double checking).
Keep UsePrimenet=1 as before.
Set the DaysOfWork to something very large, like DaysOfWork=100
Add the line SequentialWorkToDo=0
- Start Prime95 in the second directory.

Then, every day, while the second copy of Prime95 is pausing due to the Time setting:
- Edit the worktodo.ini file:
For every line that ends in ",1" (as in DoubleCheck=p,bits,1), cut the line out of worktodo.ini and paste it into the manual release form at http://www.mersenne.org/ips/manualtests.html#release. For every line that ends in ",0", just keep it in worktodo.ini. Only release exponents while the second copy of Prime95 is stopped (pausing). If you do so while it is running, it will promptly re-checkout the very same exponents that you just finished releasing.


The first copy of Prime95 keeps doing your regular work, and the second copy will specialize in P-1 testing. The Time lines ensure that the two copies of Prime95 never run simultaneously. Adjust the time periods according to your preference (but the P-1 testing can't run 24 hours a day for the reasons given below).

The second copy needs to have a large DaysOfWork value to fool the server into giving you enough exponents to P-1 test. The server thinks you're going to do a full double-check, whereas you're actually going to release the exponents after you P-1 test them.

The SequentialWorkToDo=0 line causes all the exponents in worktodo.ini to be P-1 tested before any of them is double-checked. This is described in undoc.txt.

When the second copy of Prime95 restarts each day after pausing, the first thing it will do is grab a bunch of new exponents. That's why it can't run 24/7: if it did, every time you manually released the exponents that you finished P-1 testing, Prime95 would promptly automatically re-grab those very same exponents. So you have to let it pause for a few hours at least, so that other double-checkers can grab those already-P-1-tested exponents and flush them out of the server -- during that time, your first, regular copy of Prime95 will be running. Then by the time your second copy of Prime95 starts up its day's work, the server will have a fresh batch of exponents for it.

The only non-automated part of this is once a day, when you have to remember to manually edit your worktodo.ini file and cut out the ",1" lines and paste them into the manual release form. There might be a way to automate that too, with a Perl script that automatically edits worktodo.ini and does the manual release to the server via Perl's HTTP module. Then you'd just have to set up the Perl script to run once a day at a set time (using Unix cron or equivalent Windows utility). And then the thing would be fully automated.

Maybe some of the Mersenne-aries can try this method. I've already started doing it.

Note that on average, about 150-200 double-check exponents get assigned by Primenet each day. Roughly half of those will need P-1 tests, say 75-100. Each P-1 test should take an hour or less, depending on the speed of your machine. So a very small group of people doing this type of semi-automated P-1 testing could easily clear out all the P-1 tests in the path of the leading edge of double-checking.

Last fiddled with by GP2 on 2003-09-29 at 14:59
GP2 is offline   Reply With Quote
Old 2003-12-30, 00:37   #13
GP2
 
GP2's Avatar
 
Sep 2003

22·3·5·43 Posts
Default

Updated stats (first column is the M range, eg 6 = 6M-7M range):


LL tested (verified good) without ever being P-1 trial-factored?

Code:
M   # exps
==  =====
 0      0
 1      0
 2      0
 3  22362
 4  18265
 5   9071
 6   4159
 7   1910
 8   1452
 9   2716
10    308
11     24
12      8
13      6
14      1
15      8
16      4
17      2
18      3
19      2
20      1
21      0

LL tested (unverified) without ever being P-1 trial-factored?

Code:
M   # exps
==  =====
 0      0
 1      0
 2      0
 3      0
 4      0
 5      0
 6      0
 7      0
 8     19
 9    312
10   1992
11   5055
12   3912
13   1819
14   1123
15   1030
16   1028
17   1102
18   1178
19   1429
20    889
21    167
GP2 is offline   Reply With Quote
Old 2003-12-30, 17:14   #14
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

22·3·641 Posts
Default

Let me attempt to clear up a widespread (throughout GIMPS forums, not just this thread) misconception about factoring terminology.
Quote:
P-1 trial-factoring

...

without ever being P-1 trial-factored
Trial factoring is a method of factoring. P-1 factoring is a different method of factoring. There is no such thing as "P-1 trial factoring". One can write of "trial or P-1 factoring" or "P-1 and trial factoring", but not of the two as though they were combined into a single "P-1 trial" method. Their algorithms are distinctly different. Trial factoring does not care about the factorization of some composite which is one less than some prime, but P-1 does. P-1 factoring does not "try" every possible factor up to some boundary, but trial factoring does.

Unfortunately, this distinction is not always made clear in Prime95 messages. Trial factoring was incorporated into Prime95 software long before the P-1 method was added, and so for a long time trial factoring was the only factoring method that Prime 95 used. This led to use of the single word "factoring" (or "factored") in messages that referred only to trial factoring. Such messages were not amended when the P-1 method was added, leaving a potential source of confusion.
cheesehead is offline   Reply With Quote
Old 2003-12-30, 19:01   #15
GP2
 
GP2's Avatar
 
Sep 2003

1010000101002 Posts
Default

If you say an exponent was "factored", that implies a factor was successfully found. If you say it was "trial-factored", that means factoring was attempted but not necessarily successful.

So I resisted using the phrase "P-1 factored" because it implies a factor was found. I invented the phrase "P-1 trial-factored" by analogy, but I see it may have caused even more confusion.

What's a compact adjective to indicate that an attempt at P-1 factoring was done to completion?
GP2 is offline   Reply With Quote
Old 2003-12-30, 19:02   #16
GP2
 
GP2's Avatar
 
Sep 2003

22·3·5·43 Posts
Default

Quote:
Originally posted by GP2
What's a compact adjective to indicate that an attempt at P-1 factoring was done to completion?
Hmm, maybe "P-1 tested" is the obvious answer.
GP2 is offline   Reply With Quote
Old 2003-12-30, 19:17   #17
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

8,599 Posts
Default

P-1'ed is what I say to mean that it has been tried, but no factor found (neccasarily).
Uncwilly is online now   Reply With Quote
Old 2003-12-31, 01:41   #18
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

22×3×641 Posts
Default Factoring nomenclature

Quote:
Originally posted by GP2
If you say it was "trial-factored", that means factoring was attempted but not necessarily successful.
But in the GIMPS context, George established (for better or worse) "trial factoring" as the phrase to refer to a specific factoring method.

Let's go to some authority on nomenclature.

An easy on-line reference is MathWorld. At http://mathworld.wolfram.com/PrimeFa...lgorithms.html Eric Weisstein uses direct search factorization as the name of what GIMPSters commonly call "trial factoring".

Weisstein adds "(a.k.a. trial division)" and defines the latter in a way that I can interpret to mean either an individual step of the method or the entire method.

I don't claim that Weisstein is the ultimate authority, but my practically-null offline number theory library contains no better alternative.

Shall we all agree that George made a teensy-tiny mistake in terminology and that for the good of the project a substitution shall be propogated?

Last fiddled with by cheesehead on 2003-12-31 at 01:46
cheesehead is offline   Reply With Quote
Old 2003-12-31, 05:20   #19
dsouza123
 
dsouza123's Avatar
 
Sep 2002

66210 Posts
Default

Suggestions.

For factoring methods
trial factoring
p-1 factoring

For a completed factoring test whether a factor was found or not
trial factored
p-1 factored

For a completed factoring test with a factor
successfully trial factored / trial factorable / trial factorized
successfully p-1 factored / p-1 factorable / p-1 factorized

For a completed factoring test without a factor
unsuccessfully trial factored / trial unfactorable / trial disfactorized
unsuccessfully p-1 factored / p-1 unfactorable / p-1 disfactorized
dsouza123 is offline   Reply With Quote
Old 2003-12-31, 14:49   #20
tom11784
 
tom11784's Avatar
 
Aug 2003
Upstate NY, USA

2·163 Posts
Default

my thoughts on the suggestions

For factoring methods - good
trial factoring
p-1 factoring

For a completed factoring test with a factor
trial factored
p-1 factored

For a completed factoring test without a factor
tf tested
p-1 tested
tom11784 is offline   Reply With Quote
Old 2004-03-04, 02:53   #21
markr
 
markr's Avatar
 
"Mark"
Feb 2003
Sydney

3×191 Posts
Default

Updated stats! (First column is the Millions range, eg 6 = 6M-7M)

Exponents LL tested (verified) and not P-1 tested
Code:
 M      Number
--------------
 3        3736
 4       18024
 5        8910
 6        4125
 7        1899
 8        1416
 9        2700
10         127
11          70
12           8
13           6
14           1
15           3
16           4
17           2
18           5
19           2
20           1
Exponents LL tested (unverified) and not P-1 tested
Code:
 M      Number
--------------
 8          11
 9         170
10         713
11        2307
12        3784
13        1245
14         921
15         972
16        1021
17        1106
18        1197
19        1481
20        1004
21         336
22         131
23          17
markr is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Modifying the Lucas Lehmer Primality Test into a fast test of nothing Trilo Miscellaneous Math 25 2018-03-11 23:20
Why would a website claim I've made too many requests when I haven't been back for hours? jasong jasong 5 2016-06-02 01:14
Unreserving exponents(these exponents haven't been done) jasong Marin's Mersenne-aries 7 2006-12-22 21:59
If you haven't yet ditched the Netscape browser... ewmayer Lounge 3 2005-05-10 00:28
A primality test for Fermat numbers faster than Pépin's test ? T.Rex Math 0 2004-10-26 21:37

All times are UTC. The time now is 03:31.

Wed Sep 23 03:31:55 UTC 2020 up 13 days, 42 mins, 0 users, load averages: 1.88, 1.69, 1.57

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.