mersenneforum.org  

Go Back   mersenneforum.org > Prime Search Projects > No Prime Left Behind

Reply
 
Thread Tools
Old 2008-05-02, 22:08   #12
Mini-Geek
Account Deleted
 
Mini-Geek's Avatar
 
"Tim Sorbera"
Aug 2006
San Antonio, TX USA

17·251 Posts
Default

Code:
3891584 Great Internet Mersenne Prime Search by Woltman & Kurowski
 137172 Seventeen or Bust
   5531 PrimeGrid
   5168 Riesel Sieve Project
   4414 Riesel Prime Search
   3128 The Prime Sierpinski Problem
   2771 321search
   2237 Yves Gallot's GFN Search Project
   1583 Prime Internet Eisenstein Search
   1206 GFN 2^17 Sieving project
   1019 12121 Search
    524 PrimeSearch
    324 Mat's Prime Search
    269 GFN 2^16 Sieving project
    252 Riesel Base 5
    220 15k*2^n-1 search
    162 Les GeneFermiers
     53 GFN 2^15 Sieving project
     48 GFN 2^14 Sieving project
     29 William Garnett's PSearch
Code:
PrimeGrid                        +1210
Riesel Sieve Project             +1047
Riesel Prime Search              +681
Yves Gallot's GFN Search Project +654
321search                        +535
PrimeSearch                      +29
15k*2^n-1 search                 +14
GFN 2^16 Sieving project         +1
Riesel Base 5                    -1
Seventeen or Bust                -15
GIMPS                            -431
Edit: I re-ordered the change amount list in order of changed value. I also removed the change for William Garnett's PSearch, which was not on the list previously. Unfortunately there's no easy way to show more than the top-20 projects on the top-5000 site so there's no way to know how much it or the Twin Prime Search changed since the last list. ...Gary

Last fiddled with by gd_barnes on 2008-05-22 at 10:11 Reason: Re-ordered list and last para.
Mini-Geek is offline   Reply With Quote
Old 2008-05-05, 03:32   #13
PrimeMogul
 
PrimeMogul's Avatar
 
May 2008

22 Posts
Default No change in scoring

Quote:
Originally Posted by Beyond View Post
From the Prime Pages;

The PrimeGrid/TPS project has once again uncovered a weakness in the Prime Page's submission system with their large number of submissions [snip]

looks like some changes on the way that may effect the scoring.
Nope. Just altering the way entrance rank is estimated before it is then calculated exactly. No change to the scoring is planed. Entrance rank and score are not (directly) related.
PrimeMogul is offline   Reply With Quote
Old 2008-05-07, 02:54   #14
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

5×2,017 Posts
Default

I thought I'd do an earlier update this time since we've had nice progress with the recent rally and some recent large primes:

Code:
3891584 Great Internet Mersenne Prime Search by Woltman & Kurowski
 137172 Seventeen or Bust
   5532 PrimeGrid
   5168 Riesel Sieve Project
   4460 Riesel Prime Search
   3128 The Prime Sierpinski Problem
   2771 321search
   2237 Yves Gallot's GFN Search Project
   1583 Prime Internet Eisenstein Search
   1206 GFN 2^17 Sieving project
   1031 12121 Search
    580 PrimeSearch
    324 Mat's Prime Search
    269 GFN 2^16 Sieving project
    252 Riesel Base 5
    220 15k*2^n-1 search
    162 Les GeneFermiers
     53 GFN 2^15 Sieving project
     48 GFN 2^14 Sieving project
     29 William Garnett's PSearch
Differences:
Code:
+56   PrimeSearch
+46   Riesel Prime Search
+12   12121 Search
+1    PrimeGrid
For the first time, we gained score against RPS! Nice work everyone!


Gary
gd_barnes is offline   Reply With Quote
Old 2008-05-09, 06:19   #15
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

5·2,017 Posts
Default

I decided it was time to normalize the top-5000 production scores a little bit so we could REALLY see how NPLB stacks up against other major and lesser projects. For every top-20 project on the top-5000 site, what I did was took the production score on each and divided it by the number of years since the project found it's first prime. Some projects are now defunct so I used the number of years between the first and last prime that was found.

In order to properly compute our score, I did a detailed analysis of the production score that the PrimeSearch project had before NPLB was started and the score since it was started. Prior to 110 days ago when NPLB started, all primes for Primesearch had a total combined production score of 215 in over 9 YEARS! Since NPLB took over the effort, we have had a total production score of 375 in just 3-1/2 MONTHS!! (Current total 590.)

For an analysis of all projects, see the attached spreadsheet. What I concluded was quite satisfying. Here's the bottom line:

Primesearch before NPLB: it was the WORST project of any of the top-20 projects on the top-5000 site with a score of just 23 / year. It had truly languished in obscurity for an extended period.

-BUT-

Since NPLB took over the effort, it has been the 5TH BEST project on the site with a score of 1286 / year. We have certainly made a splash on the prime searching community.

This beats even the (presumably) mammonth RieselSieve project that is averaging 1170 / year even after finding a recent huge prime.

We are just a little ways behind RPS that is averaging 1727 / year.

Starting from scratch, this is a HUGE accomplishment in just 3.5 months! I/we owe a debt of gratitude to our faithful searchers. We couldn't have done it without you!

What we need now is about 35-40% more total computing power on our ranges and perhaps we can finally close the gap with RPS in production score per month/year. This is not a huge gap to fill so if anyone knows of anyone else that would like to get involved with us, bring them our way.


Gary
Attached Files
File Type: zip Project score per year.zip (2.7 KB, 72 views)
gd_barnes is offline   Reply With Quote
Old 2008-05-10, 02:15   #16
IronBits
I ♥ BOINC!
 
IronBits's Avatar
 
Oct 2002
Glendale, AZ. (USA)

100010110012 Posts
Default

Is any project doing "double work" ?
In other words, is any other project doing the same work, thus doing the same work twice?
IronBits is offline   Reply With Quote
Old 2008-05-13, 05:08   #17
IronBits
I ♥ BOINC!
 
IronBits's Avatar
 
Oct 2002
Glendale, AZ. (USA)

3·7·53 Posts
Default

Bump. Anyone?
IronBits is offline   Reply With Quote
Old 2008-05-13, 05:29   #18
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

3·2,083 Posts
Default

Quote:
Originally Posted by IronBits View Post
Is any project doing "double work" ?
In other words, is any other project doing the same work, thus doing the same work twice?
Quote:
Originally Posted by IronBits View Post
Bump. Anyone?
Short answer: mostly no.

Long answer: Most projects are searching different ranges, or even different types, of numbers, than each other, so there is little double work between projects. Projects usually don't even want to search other projects' ranges--it would only mean lots of double work and lots of searchers being disappointed by finding a prime, only to find that someone else found it first.

Gary can probably give a fuller answer next time he gets on (he's on another business trip right now, so he isn't able to get on as often as usual).

Anon
mdettweiler is offline   Reply With Quote
Old 2008-05-13, 17:40   #19
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

235458 Posts
Default

I concur with everything that Anon said.

Projects attempt to avoid other projects efforts. We are, in effect, double-checking some INDIVIDUAL's efforts that were previously mostly done at Primesearch and a little bit at RPS but that is mostly because we've found so many problems with the reporting of Primesearch primes. Our effort probably has more overlap with any prior effort than just about any project but it was a necessity to have 'no prime left behind'.

In stating this, up until NPLB, Primesearch had just been a lot of fragmented individual efforts with little coordination, which is why the "project" languished for so long.


Gary
gd_barnes is offline   Reply With Quote
Old 2008-05-21, 08:29   #20
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

276516 Posts
Default

With help from Anon and Carlos both finding primes for n>600K, we were tops again! Nice work everyone! It looks like the n=333333 primes dropping off finally had a major negative impact on all of the scores. NPLB, CRUS, and RPS were the only ones that gained score.

Code:
3868630 Great Internet Mersenne Prime Search by Woltman & Kurowski
 136364 Seventeen or Bust
   5507 PrimeGrid
   5138 Riesel Sieve Project
   4526 Riesel Prime Search
   3110 The Prime Sierpinski Problem
   2755 321search
   2224 Yves Gallot's GFN Search Project
   1554 Prime Internet Eisenstein Search
   1199 GFN 2^17 Sieving project
   1025 12121 Search
    648 PrimeSearch
    322 Mat's Prime Search
    268 GFN 2^16 Sieving project
    250 Riesel Base 5
    217 15k*2^n-1 search
    154 Les GeneFermiers
     66 Conjectures 'R Us
     53 GFN 2^15 Sieving project
     48 GFN 2^14 Sieving project
Differences:
Code:
   +68 PrimeSearch
   +66 Riesel Prime Search
   +40 Conjectures 'R Us
    -1 GFN 2^16 Sieving project
    -2 Mat's Prime Search
    -2 Riesel Base 5
    -3 15k*2^n-1 search
    -6 12121 Search
    -7 GFN 2^17 Sieving project
    -8 Les GeneFermiers
   -13 Yves Gallot's GFN Search Project
   -16 321search
   -18 The Prime Sierpinski Problem
   -25 PrimeGrid
   -29 Prime Internet Eisenstein Search
   -30 Riesel Sieve Project
  -808 Seventeen or Bust
-22954 Great Internet Mersenne Prime Search by Woltman & Kurowski

Last fiddled with by gd_barnes on 2008-05-21 at 08:43
gd_barnes is offline   Reply With Quote
Old 2008-05-21, 12:09   #21
Mini-Geek
Account Deleted
 
Mini-Geek's Avatar
 
"Tim Sorbera"
Aug 2006
San Antonio, TX USA

426710 Posts
Default

Quote:
Originally Posted by gd_barnes View Post
With help from Anon and Carlos both finding primes for n>600K, we were tops again! Nice work everyone! It looks like the n=333333 primes dropping off finally had a major negative impact on all of the scores. NPLB, CRUS, and RPS were the only ones that gained score.

Code:
3868630 Great Internet Mersenne Prime Search by Woltman & Kurowski
 136364 Seventeen or Bust
   5507 PrimeGrid
   5138 Riesel Sieve Project
   4526 Riesel Prime Search
   3110 The Prime Sierpinski Problem
   2755 321search
   2224 Yves Gallot's GFN Search Project
   1554 Prime Internet Eisenstein Search
   1199 GFN 2^17 Sieving project
   1025 12121 Search
    648 PrimeSearch
    322 Mat's Prime Search
    268 GFN 2^16 Sieving project
    250 Riesel Base 5
    217 15k*2^n-1 search
    154 Les GeneFermiers
     66 Conjectures 'R Us
     53 GFN 2^15 Sieving project
     48 GFN 2^14 Sieving project
Differences:
Code:
   +68 PrimeSearch
   +66 Riesel Prime Search
   +40 Conjectures 'R Us
    -1 GFN 2^16 Sieving project
    -2 Mat's Prime Search
    -2 Riesel Base 5
    -3 15k*2^n-1 search
    -6 12121 Search
    -7 GFN 2^17 Sieving project
    -8 Les GeneFermiers
   -13 Yves Gallot's GFN Search Project
   -16 321search
   -18 The Prime Sierpinski Problem
   -25 PrimeGrid
   -29 Prime Internet Eisenstein Search
   -30 Riesel Sieve Project
  -808 Seventeen or Bust
-22954 Great Internet Mersenne Prime Search by Woltman & Kurowski
That's cool that NPLB is the project behind the largest gain. Two of your projects are in the three projects that gained.
What caused the huge losses for GIMPS and SoB? For GIMPS was it that GIMPS's first prime got pushed off the top 100 recently? I thought these scores were for anything still on the top 5000.
Edit: Post 600

Last fiddled with by Mini-Geek on 2008-05-21 at 12:13
Mini-Geek is offline   Reply With Quote
Old 2008-05-21, 15:12   #22
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

624910 Posts
Default

Quote:
Originally Posted by Mini-Geek View Post
That's cool that NPLB is the project behind the largest gain. Two of your projects are in the three projects that gained.
What caused the huge losses for GIMPS and SoB? For GIMPS was it that GIMPS's first prime got pushed off the top 100 recently? I thought these scores were for anything still on the top 5000.
Edit: Post 600
The normalized scores, according to the top-5000 web site, are normalized based on the current 5000th-place prime, thus if a project doesn't keep on finding primes at a decent clip, its score will steadily drop. (Similar to Gary's score-by-year Excel spreadsheet he posted earlier.) With n=333333 primes monopolizing the list so long, though, that "dropping effect" was neutralized. Now that they're all off the list, we see that effect again.

As for why GIMPS is dropping so much in comparison to the others...I guess it's just because their primes are so, so, so big, that the "dropping effect" appears much larger. The normalized scores will drop by an amount proportional to the amount of normalized score the project has.
mdettweiler is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Best msieve poly scores VBCurtis Msieve 160 2020-03-13 12:19
4000 < k < 5000 otutusaus Riesel Prime Data Collecting (k*2^n-1) 5 2012-03-07 20:01
5000 < k < 6000 justinsane Riesel Prime Data Collecting (k*2^n-1) 26 2010-12-31 12:27
As Math Scores Lag, a New Push for the Basics -- New York Times ewmayer Soap Box 14 2006-11-17 09:40
Top-5000 List edorajh Riesel Prime Search 17 2006-03-28 21:57

All times are UTC. The time now is 21:58.

Sun Mar 29 21:59:02 UTC 2020 up 4 days, 19:32, 2 users, load averages: 1.15, 1.36, 1.40

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.