mersenneforum.org  

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

Reply
 
Thread Tools
Old 2010-12-16, 12:59   #1
Mr. P-1
 
Mr. P-1's Avatar
 
Jun 2003

22218 Posts
Default Insufficient information for accurate CPU credit.

I just submitted the following two factors via the manual testing form:

Code:
[Fri Nov 26 10:38:45 2010]
P-1 found a factor in stage #1, B1=515000.
UID: daran/cuica, M42785489 has a factor: 340930634243704316351, AID: 4B29AFF5F9D1EB7A3C61A9B9118851B9
[Thu Dec 16 08:06:37 2010]
P-1 found a factor in stage #1, B1=465000.
UID: daran/cuica, M39394703 has a factor: 51759351545613783550417, AID: 460389DCB57262419281C8E7E4C6AD36
Here's the response in full:

Code:
No factor lines found: 0
Mfaktc no factor lines found: 0
Factors found: 2
Processing result: M42785489 has a factor: 340930634243704316351, AID: 4B29AFF5F9D1EB7A3C61A9B9118851B9
Insufficient information for accurate CPU credit.
For stats purposes, assuming factor was found using P-1 with B1 = 800000.
CPU credit is 2.0139 GHz-days.
Processing result: M39394703 has a factor: 51759351545613783550417, AID: 460389DCB57262419281C8E7E4C6AD36
Insufficient information for accurate CPU credit.
For stats purposes, assuming factor was found using P-1 with B1 = 800000.
CPU credit is 1.5273 GHz-days.
P-1 lines found: 0
LL lines found: 0
Mlucas lines found: 0
Glucas (G29) lines found: 0
Glucas lines found: 0
MacLucasFFTW lines found: 0
CUDALucas lines found: 0
ECM lines found: 0
I'm grateful for the generous credit, but I don't see why it couldn't give me the correct amount.
Mr. P-1 is offline   Reply With Quote
Old 2010-12-17, 01:16   #2
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

22×3×641 Posts
Default

Quote:
Originally Posted by Mr. P-1 View Post
I just submitted the following two factors via the manual testing form:

Code:
[Fri Nov 26 10:38:45 2010]
P-1 found a factor in stage #1, B1=515000.
UID: daran/cuica, M42785489 has a factor: 340930634243704316351, AID: 4B29AFF5F9D1EB7A3C61A9B9118851B9
[Thu Dec 16 08:06:37 2010]
P-1 found a factor in stage #1, B1=465000.
UID: daran/cuica, M39394703 has a factor: 51759351545613783550417, AID: 460389DCB57262419281C8E7E4C6AD36
So, you copied text from results.txt into the manual report form, right?

Quote:
Here's the response in full:

Code:
No factor lines found: 0
Mfaktc no factor lines found: 0
Factors found: 2
Processing result: M42785489 has a factor: 340930634243704316351, AID: 4B29AFF5F9D1EB7A3C61A9B9118851B9
Insufficient information for accurate CPU credit.
For stats purposes, assuming factor was found using P-1 with B1 = 800000.
CPU credit is 2.0139 GHz-days.
Processing result: M39394703 has a factor: 51759351545613783550417, AID: 460389DCB57262419281C8E7E4C6AD36
Insufficient information for accurate CPU credit.
For stats purposes, assuming factor was found using P-1 with B1 = 800000.
CPU credit is 1.5273 GHz-days.
P-1 lines found: 0
LL lines found: 0
Mlucas lines found: 0
Glucas (G29) lines found: 0
Glucas lines found: 0
MacLucasFFTW lines found: 0
CUDALucas lines found: 0
ECM lines found: 0
I'm grateful for the generous credit, but I don't see why it couldn't give me the correct amount.
Theory:

IIRC (it's been a while since I did this, and I can't find a saved example) the automatic report text to PrimeNet does not include the actual B1/B2 when reporting a factor. So, PrimeNet can't calculate credit on that basis, but it generates a B1 based on some formula.

Your manual report included B1(/B2), but PrimeNet is ignoring that and performing the credit calculation without actual B1/B2, but generating a B1 based on that same formula, just as it does for the automatic reports.

Last fiddled with by cheesehead on 2010-12-17 at 01:19
cheesehead is offline   Reply With Quote
Old 2010-12-17, 01:53   #3
axn
 
axn's Avatar
 
Jun 2003

32·19·29 Posts
Default

Quote:
Originally Posted by cheesehead View Post
IIRC (it's been a while since I did this, and I can't find a saved example) the automatic report text to PrimeNet does not include the actual B1/B2 when reporting a factor. So, PrimeNet can't calculate credit on that basis, but it generates a B1 based on some formula.

Your manual report included B1(/B2), but PrimeNet is ignoring that and performing the credit calculation without actual B1/B2, but generating a B1 based on that same formula, just as it does for the automatic reports.
You've got it backwards. Automatic report includes B1/B2 and primenet calculates accurate credit. Results.txt does not contain B1/B2. So primenet uses some default calculations. Note that the "B1=" text is not in the same line as the result and so the manual credit page is not able to associate one with the other.
axn is offline   Reply With Quote
Old 2010-12-17, 02:14   #4
Mr. P-1
 
Mr. P-1's Avatar
 
Jun 2003

7×167 Posts
Default

Quote:
Originally Posted by cheesehead View Post
So, you copied text from results.txt into the manual report form, right?
Yes.


Quote:
Theory:

IIRC (it's been a while since I did this, and I can't find a saved example) the automatic report text to PrimeNet does not include the actual B1/B2 when reporting a factor. So, PrimeNet can't calculate credit on that basis, but it generates a B1 based on some formula.
I don't think so. Here are some of my results reported automatically:

Code:
cuica	39046153	NF-PM1	2010-08-08 17:19	93.7	B1=465000, B2=12090000	 2.0970
cuica	39008107	NF-PM1	2010-08-08 17:19	28.8	B1=465000, B2=12090000	2.0970
cuica	39008779	F-PM1	2010-08-08 17:19	28.8	6397283804036361597329	2.0970
cuica	39968749	NF-PM1	2010-08-08 17:19	0.0	B1=480000, B2=12360000	2.8377
cuica	39973847	NF-PM1	2010-08-05 07:41	90.3	B1=480000, B2=12360000	2.8377
cuica	39973279	F-PM1	2010-08-04 23:14	90.0	4632811025542760339833	1.2083
cuica	39007921	NF-PM1	2010-08-04 07:11	24.3	B1=465000, B2=11857500	2.0728
cuica	39969173	NF-PM1	2010-08-03 15:10	88.6	B1=480000, B2=12360000	2.8377
cuica	39965917	NF-PM1	2010-08-01 11:53	86.5	B1=480000, B2=12360000	2.8377


cuica	39015863	NF-PM1	2010-07-12 08:26	66.3	B1=465000, B2=12090000	 2.0970
cuica	39012313	NF-PM1	2010-07-11 14:26	65.6	B1=465000, B2=12090000	2.0970
cuica	39010459	F-PM1	2010-07-10 18:58	64.8	1097229813650567416577	2.2539
cuica	39010901	NF-PM1	2010-07-09 22:18	63.9	B1=465000, B2=12090000	2.0970
cuica	39011341	NF-PM1	2010-07-08 23:50	63.0	B1=465000, B2=11857500	2.0728
cuica	39010603	NF-PM1	2010-07-08 06:23	62.2	B1=465000, B2=11857500	2.0728
cuica	39740093	F-PM1	2010-07-07 12:33	21.1	218423414608878788302287871	2.8081
cuica	39737561	NF-PM1	2010-07-06 12:02	20.1	B1=475000, B2=12231250	2.8081
cuica	39009119	NF-PM1	2010-07-05 12:13	19.1	B1=465000, B2=11857500	2.0728
cuica	39691237	NF-PM1	2010-07-04 18:12	18.3	B1=475000, B2=12231250	2.8081
cuica	39682717	F-PM1	2010-07-03 18:32	17.3	624302998170670921801	1.1957
cuica	40420057	NF-PM1	2010-07-03 16:18	56.6	B1=455000, B2=10806250	2.5651
cuica	39007183	NF-PM1	2010-07-02 16:28	16.2	B1=465000, B2=12090000	2.0970
cuica	39005611	NF-PM1	2010-07-01 20:12	15.4	B1=465000, B2=12090000	2.0970
cuica	38988233	NF-PM1	2010-07-01 00:43	14.6	B1=465000, B2=12090000	2.0970
cuica	39007219	NF-PM1	2010-06-30 05:15	13.8	B1=465000, B2=12090000	2.0970
Note that in two cases, a stage 2 factor received identical credit to that for comparable unsuccessful attempts. And this, in general, is what I'm used to seeing. Stage 1 factors get a little under half the credit, again consistent with my observation that the program spends a little under its time in stage 1.

The one anomaly is the 7/10 factor, whose credit is not matched or even close to any other result, successful or not, in the past year. The most likely explanation is that this exponent was assigned with a much lower TF level, and consequently was P-1ed. to higher bounds.

In the case of my manual reports, the server did compute its own bound, one that it considerably at odds with the bounds typically used for exponents of this size:
Mr. P-1 is offline   Reply With Quote
Old 2010-12-17, 02:23   #5
Mr. P-1
 
Mr. P-1's Avatar
 
Jun 2003

7·167 Posts
Default

Quote:
Originally Posted by axn View Post
You've got it backwards. Automatic report includes B1/B2 and primenet calculates accurate credit. Results.txt does not contain B1/B2. So primenet uses some default calculations. Note that the "B1=" text is not in the same line as the result and so the manual credit page is not able to associate one with the other.
Clearly results.txt does contain B1/B2, but not on the same line, as you say, so that may be the reason.

It's not unreasonable for Primenet to ignore the bounds in manual reports, regardless of where they appear, on the grounds that they're easily forged. That said, the current system is open to abuse: I could submit all my stage 1 factors manually to gain extra credit, while allowing my stage 2 factors to be reported in the normal way.
Mr. P-1 is offline   Reply With Quote
Old 2010-12-17, 03:49   #6
axn
 
axn's Avatar
 
Jun 2003

32×19×29 Posts
Default

Quote:
Originally Posted by Mr. P-1 View Post
Clearly results.txt does contain B1/B2, but not on the same line, as you say, so that may be the reason.

It's not unreasonable for Primenet to ignore the bounds in manual reports, regardless of where they appear, on the grounds that they're easily forged. That said, the current system is open to abuse: I could submit all my stage 1 factors manually to gain extra credit, while allowing my stage 2 factors to be reported in the normal way.
considering how few factors are found by p-1 stage 1, good luck with that!
axn is offline   Reply With Quote
Old 2010-12-17, 20:29   #7
Mini-Geek
Account Deleted
 
Mini-Geek's Avatar
 
"Tim Sorbera"
Aug 2006
San Antonio, TX USA

17×251 Posts
Default

Quote:
Originally Posted by axn View Post
considering how few factors are found by p-1 stage 1, good luck with that!
Assuming everything works just like the one test I just did (with 400 MB allowed; and I'm sure it doesn't work perfectly the same, but probably close enough for these purposes), and Prime95's "Chance of finding a factor..." is accurate, and the difference between (expected factors in stage 1)/(expected factors total) is not significantly different from (chance of factor in stage 1)/(chance of factor total) about 47% of P-1 factors in two-stage tests are found by stage 1. But still, P-1 factors are rare enough, and the difference in credit small enough, that if you abused this to the maximum potential, you'd probably only get about 1.2% more credit on average than you deserve.
Mini-Geek is online now   Reply With Quote
Old 2010-12-17, 21:56   #8
Bdot
 
Bdot's Avatar
 
Nov 2010
Germany

59710 Posts
Default

Quote:
Originally Posted by Mini-Geek View Post
... you'd probably only get about 1.2% more credit on average than you deserve.
I usually get way less credit than really spent when manually reporting:

Quote:
Manual testing 53470633 NF-PM1 2010-11-29 08:55 4.7 B1=635000, B2=18097500 4.8753
Manual testing 53470619 F-PM1 2010-11-29 08:55 4.7 1919565392651028059642113943994258791 2.4586
This means when a factor is found, only half the credit of NF-PM1 is granted. The amount of work was the same as the GCD of stage2 found it.

And this is even worse when finding a factor in ECM:
Quote:
Originally Posted by results.txt
[Fri Dec 17 20:01:14 2010]
ECM found a factor in curve #85, stage #2
Sigma=4045519484526505, B1=50000, B2=5000000.
M5392193 has a factor: 9125624223117137685847, AID: A0B3D8BDD38BE54BB3F5E0A96196011F
results in
Quote:
Originally Posted by response
Factors found: 1
Processing result: M5392193 has a factor: 9125624223117137685847, AID: A0B3D8BDD38BE54BB3F5E0A96196011F
Insufficient information for accurate CPU credit.
For stats purposes, assuming factor was found using ECM with B1 = 50000.
CPU credit is 0.2695 GHz-days.
The first 84 curves with no factor are ignored ... This test took almost a month (on a rather slow machine, still it should have been somewhere near 20 GHz-days).

Maybe it's time to improve the manual reporting to not ignore the lines before the "has a factor" one? The information for accurate accounting is there ...
Bdot is offline   Reply With Quote
Old 2010-12-17, 22:09   #9
axn
 
axn's Avatar
 
Jun 2003

115378 Posts
Thumbs up

Quote:
Originally Posted by Mini-Geek View Post
But still, P-1 factors are rare enough, and the difference in credit small enough, that if you abused this to the maximum potential, you'd probably only get about 1.2% more credit on average than you deserve.
Sounds about right.
axn is offline   Reply With Quote
Old 2010-12-17, 22:15   #10
axn
 
axn's Avatar
 
Jun 2003

32·19·29 Posts
Default

Quote:
Originally Posted by Bdot View Post
Maybe it's time to improve the manual reporting to not ignore the lines before the "has a factor" one? The information for accurate accounting is there ...
Unless that line itself is checksum protected to prevent/detect tampering, I can hand edit the values to something ridiculously big*. Now _that_ would be open to abuse.

*Find the largest factor of p-1. Set my B1 = that factor - 1. Then set B2 = 10000 * that factor.
For eg:- 9125624223117137685847-1 = 2*3*13*109*5392193*199056357361
Set B1 = 199056357360 and B2 = 1990563573600000. Watch the credits roll in
axn is offline   Reply With Quote
Old 2010-12-18, 03:41   #11
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

22×3×641 Posts
Default

Quote:
Originally Posted by axn View Post
You've got it backwards. Automatic report includes B1/B2 and primenet calculates accurate credit.
Thanks for the correction. Now I can recall the time when I (automatically) sent multiple closely-spaced-B1 reports for an exponent, and as a result got way, way more credit than deserved because each credit assumed I had started at zero.
cheesehead is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Insufficient Information for accurate CPU Credit TheMawn GPU Computing 4 2014-06-07 20:52
Account Information Primeinator Information & Answers 13 2011-12-03 19:42
Insufficient information for accurate CPU credit Yura Information & Answers 11 2011-06-22 11:52
Program information use Unregistered Information & Answers 0 2009-08-30 19:03
Is More Information Better? davar55 Puzzles 3 2009-07-02 20:25

All times are UTC. The time now is 15:57.

Sat May 8 15:58:00 UTC 2021 up 30 days, 10:38, 1 user, load averages: 3.31, 3.28, 3.27

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.