Double-Check LL (deprecated) - Page 2

Andrew Usher · 2023-02-15, 00:23

If that really is true - and you need to go by the credit actually awarded, not anything the server says about reduced credit - then it has the appearance of a bug, and it could be reported as such.

In that case you could try to do your manual DCs without an assignment - results are always accepted.

1997rj7 · 2023-02-15, 02:31

Here are screenshots from my Work Results Details Report.

Speaking of manual double checks, the new way (Double-check using PRP with proof), has its own problems. For some reason it gives out Category 0 exponents. It says they will expire in 7 days, but when 7 days comes around, it says you now have 23 more days. And when 30 days have elapsed, it says it will expire in negative days. I don't know when or if it will actually expire.

Uncwilly · 2023-02-15, 02:59

Quote:

Originally Posted by 1997rj7

For some reason it gives out Category 0 exponents. It says they will expire in 7 days, but when 7 days comes around, it says you now have 23 more days.

You have 7 days to report that you have started the exponent.

Quote:

Originally Posted by https://www.mersenne.org/thresholds/

Assignments are recycled if assignment is not started with 7 days or when assignment is more than 30 days old.

That is for First Time Checks.

The server does not handle PRP-DC correctly. They should be treated as DC's and those assignment rules should apply. But, it treats all PRP (including PRP-CF) as if it is a FTC.

Andrew Usher · 2023-02-16, 00:01

1997rj7: The screenshots are sufficient proof for me. That discrepancy is presumably not currently intended (and I don't know how it could arise), and could rightly be posted in the server problems thread, although no one is responding there, either, at the current time.

To amplify the point in the previous post, those assignments will not actually expire at the time stated (this has been discussed recently) - treating them as though they were LL is probably safe.

kdelisle2005 · 2023-02-21, 13:58

I just loaded results and they are accepted as is with full credit. I have other PCs with manual entries that may produce a different result but it will be another 10 days.

~~

Output when results were uploaded.

Found 5 lines to process at 2023-02-21T13:46:06
processing: PRP (not-prime) for M63829771
Result type (150=PRP_COMPOSITE) appropriate for the assignment type (150=PRP). Processing result and deleting assignment.
CPU credit is 149.9446 GHz-days.

processing: LL for M63829817
Result type (100=LL_COMPOSITE) appropriate for the assignment type (101=DBLCHK). Processing result and deleting assignment.
LL test successfully completes double-check of M63829817 -- CPU credit is 149.9447 GHz-days.

kdelisle2005 · 2023-02-21, 14:21

Thank you 1997rj7 for your attached results screenshot.

I wanted to capture the results message that curtailed the credits to show the action as it happened.

The below are my results that have been discounted to 1/3 credit - there are more than I thought.

Click image for larger version

Name: p95results.JPG
Views: 48
Size: 499.3 KB
ID: 28047

Obviously when this 1/3 credit issue started I was not paying attention to my uploads. Then I started to assume that someone implemented a different way to calculate the credit.

Seems like someone could query the results database and re-compute the true earned credits :)

Andrew Usher · 2023-02-23, 01:10

Evidently at some point the credit was reduced, intentionally or not, and then later this was corrected. As I have stated before, I only care about credit as it allows me to evaluate my own computing effort contributed, and for that purpose my personal summary should not apply any penalties whatever.

kdelisle2005 · 2023-03-03, 17:09

Manual testing earned credit issue appears to be fixed.

Well when I started this thread about reduced earned credit, I should have captured the screen out but did not :(

I have uploaded all of my manual results and each one posted with full credit earned. So the issue where someone altered the code which calculates the earned credit has been resolved. thank you for all double-checkers going forward.

However what can be done with my 56 results that were uploaded prior to the fix? I'm sure there are others as well that were impacted by this earned credit issue, how can others be made whole as well??

My permissions do not allow me to query other work assignments that may have been impacted but I would think someone with back-office authority can.

My guess is more than likely the credits will not be properly updated retroactively :(

Andrew Usher · 2023-03-04, 13:38

I don't think credit is ever changed retroactively, unfortunately. The word 'deprecated' should still be removed for the reasons given.

So if they ever get around to fixing the excessive credit with the new P-1 and ECM stage 2, that presumably won't be retroactive either, and we'll still see ridiculous GHz-day values in the history - the longest P-1 runs ever done have been a few thousand 'real' GHz-days, but there are very few above 10 on sub-wavefront exponents.

2023-02-15, 02:31	#13
1997rj7 Sep 2003 546₈ Posts	Here are screenshots from my Work Results Details Report. Speaking of manual double checks, the new way (Double-check using PRP with proof), has its own problems. For some reason it gives out Category 0 exponents. It says they will expire in 7 days, but when 7 days comes around, it says you now have 23 more days. And when 30 days have elapsed, it says it will expire in negative days. I don't know when or if it will actually expire. Attached Thumbnails

2023-02-21, 13:58	#16
kdelisle2005 Mar 2006 Michigan 2⁴ Posts	I just loaded results and they are accepted as is with full credit. I have other PCs with manual entries that may produce a different result but it will be another 10 days. ~~ Output when results were uploaded. Found 5 lines to process at 2023-02-21T13:46:06 processing: PRP (not-prime) for M63829771 Result type (150=PRP_COMPOSITE) appropriate for the assignment type (150=PRP). Processing result and deleting assignment. CPU credit is 149.9446 GHz-days. processing: LL for M63829817 Result type (100=LL_COMPOSITE) appropriate for the assignment type (101=DBLCHK). Processing result and deleting assignment. LL test successfully completes double-check of M63829817 -- CPU credit is 149.9447 GHz-days. Last fiddled with by kdelisle2005 on 2023-02-21 at 13:59

2023-02-21, 14:21	#17
kdelisle2005 Mar 2006 Michigan 10₁₆ Posts	Thank you 1997rj7 for your attached results screenshot. I wanted to capture the results message that curtailed the credits to show the action as it happened. The below are my results that have been discounted to 1/3 credit - there are more than I thought. Obviously when this 1/3 credit issue started I was not paying attention to my uploads. Then I started to assume that someone implemented a different way to calculate the credit. Seems like someone could query the results database and re-compute the true earned credits :) Last fiddled with by kdelisle2005 on 2023-02-21 at 14:27

2023-03-03, 17:09	#19
kdelisle2005 Mar 2006 Michigan 2⁴ Posts	Manual testing earned credit issue appears to be fixed. Well when I started this thread about reduced earned credit, I should have captured the screen out but did not :( I have uploaded all of my manual results and each one posted with full credit earned. So the issue where someone altered the code which calculates the earned credit has been resolved. thank you for all double-checkers going forward. However what can be done with my 56 results that were uploaded prior to the fix? I'm sure there are others as well that were impacted by this earned credit issue, how can others be made whole as well?? My permissions do not allow me to query other work assignments that may have been impacted but I would think someone with back-office authority can. My guess is more than likely the credits will not be properly updated retroactively :( Last fiddled with by kdelisle2005 on 2023-03-03 at 17:14 Reason: presentation

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Double-Check	wombatman	Conjectures 'R Us	3	2016-08-29 20:46
Double Check	Unregistered	Information & Answers	3	2011-10-01 04:38
First check and double check llrnet servers.	opyrt	Prime Sierpinski Project	3	2009-01-02 01:50
Double-check check?	M0CZY	Software	15	2008-10-30 14:20
Double Check P-1	PhilF	PrimeNet	6	2005-07-03 14:36

2023-02-15, 00:23	#12
Andrew Usher Dec 2022 111000111₂ Posts	If that really is true - and you need to go by the credit actually awarded, not anything the server says about reduced credit - then it has the appearance of a bug, and it could be reported as such. In that case you could try to do your manual DCs without an assignment - results are always accepted.

2023-02-16, 00:01	#15
Andrew Usher Dec 2022 1C7₁₆ Posts	1997rj7: The screenshots are sufficient proof for me. That discrepancy is presumably not currently intended (and I don't know how it could arise), and could rightly be posted in the server problems thread, although no one is responding there, either, at the current time. To amplify the point in the previous post, those assignments will not actually expire at the time stated (this has been discussed recently) - treating them as though they were LL is probably safe.

2023-02-23, 01:10	#18
Andrew Usher Dec 2022 5×7×13 Posts	Evidently at some point the credit was reduced, intentionally or not, and then later this was corrected. As I have stated before, I only care about credit as it allows me to evaluate my own computing effort contributed, and for that purpose my personal summary should not apply any penalties whatever.

2023-03-04, 13:38	#20
Andrew Usher Dec 2022 707₈ Posts	I don't think credit is ever changed retroactively, unfortunately. The word 'deprecated' should still be removed for the reasons given. So if they ever get around to fixing the excessive credit with the new P-1 and ECM stage 2, that presumably won't be retroactive either, and we'll still see ridiculous GHz-day values in the history - the longest P-1 runs ever done have been a few thousand 'real' GHz-days, but there are very few above 10 on sub-wavefront exponents.