| Author |
Message |
|---|
|
|
|
What causes a SOB wu to get as result : inconclusive ?
Completed yes but validation inconclusive
Is to due to (failure in the) algo used or ?
Can it be an indication a prime was found but the wu needs a re-check?
Even after the doublechecker gets the same result (validation inconclusive)?
For instance :
http://www.primegrid.com/result.php?resultid=739489606
Name llrSOB_243719155_1
Workunit 496252229
Outcome Success
Validate state Checked, but no consensus yet
Credit 0.00
Application version Seventeen or Bust v7.06 |
|
|
 RafaelVolunteer tester
Send message
Joined: 22 Oct 14
Posts: 911
ID: 370496
Credit: 550,234,793
RAC: 442,833
                         
|
What causes a SOB wu to get as result : inconclusive ?
Completed yes but validation inconclusive
Is to due to (failure in the) algo used or ?
In general Boinc terms, you say that the result is "A" where the other person said it was "B". So the server can't decide (and with good reason) which of oyu is correct, and thus marks it as "Inconclusive", sending a 3rd task to be crunched and be the tie breaker. If the 3rd tasks returns either A or B, then the server will know who did the calculation properly and will award that person with a regular "Valid" WU while telling the wrong guy that his result was "Invalid".
Looking at your wingman, that computer has 4 tasks which turned out with wrong results. Most likely, you did your calculation correctly and he's done it wrong, so don't sweat about it. Give it some time for the 3rd computer to crunch it and it'll probably get validated just fine. |
|
|
JimB Honorary cruncher
Send message
Joined: 4 Aug 11
Posts: 920
ID: 107307
Credit: 989,246,873
RAC: 201,706
               
|
|
The third computer will almost certainly time out in four days. It doesn't appear to have been working on the job at all (no trickles). As far as your current wingman, when you look at his workunits that didn't validate, it's all validate errors. He didn't upload anything at the end of those four jobs, so there was nothing to compare. That's what a validate error is, something is wrong with the job such that we didn't even try to compare the residue. Neither of the two current results looks obviously bad, only a third result (from a fourth or later host) can determine which one is correct. I've had these jobs wait six months for a wingman to finish. It's better these days, but don't hold your breath while waiting. |
|
|
|
|
|
Sorry for a random bump. It's my first time crunching SOB tasks...
Workunit 502773035
Is it common for SOB to have 4+ returned result with no consensus? They all looks returned with reasonable CPU time and the stderr looks normal :\
What would happen if we keep getting results that failed to give an ultimate agreement? |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 433,172,407
RAC: 1,010,059
                  
|
Sorry for a random bump. It's my first time crunching SOB tasks...
Workunit 502773035
Is it common for SOB to have 4+ returned result with no consensus? They all looks returned with reasonable CPU time and the stderr looks normal :\
What would happen if we keep getting results that failed to give an ultimate agreement?
No, it's definitely not common, but it's not unprecidented either. I just looked at the results, and all 4 are different. At least three of them are wrong, of course
The second and third results are from computers with a history of producing bad results, so those two are probably wrong. That leaves your result and the first result. At least one is wrong, but you'll have to wait for another result to find out which one is wrong.
The bad results are because of computation errors in the computers. There's nothing wrong with the task itself. You'll eventually get a consensus once two computers return the correct result.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
If mine turned out to be wrong should I consider giving up crunching SoB?
The result was from a laptop, but since LLR-tasks only take a small portion of my crunching work I managed to keep the temperature reasonably safe (75-80). All other primegrid tasks worked fine on this machine but I am just not sure about SoB since this is the longest task I have crunched other than CPDN.
Anyway thanks for answering my question. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 433,172,407
RAC: 1,010,059
|
If mine turned out to be wrong should I consider giving up crunching SoB?
The result was from a laptop, but since LLR-tasks only take a small portion of my crunching work I managed to keep the temperature reasonably safe (75-80). All other primegrid tasks worked fine on this machine but I am just not sure about SoB since this is the longest task I have crunched other than CPDN.
Anyway thanks for answering my question.
It means your computer is not working correctly. This is a problem that will only get worse over time. What you should do is fix the computer.
____________
My lucky number is 75898524288+1 |
|
|
|
|
That leaves your result and the first result. At least one is wrong, but you'll have to wait for another result to find out which one is wrong.
And UID 193455 has done one SOB (feb7). And it's failed. What? Show must go on ;)
____________
I'm counting for science,
Points just make me sick. |
|
|
JimBHonorary cruncher
Send message
Joined: 4 Aug 11
Posts: 920
ID: 107307
Credit: 989,246,873
RAC: 201,706
|
|
I manually ran that workunit here and inserted the result. Don't read anything into the ratio between CPU and elapsed time because CPU = elapsed * 4 since I used four cores and all I had was the elapsed time from the LLR outcome string. I did this because I don't really expect the guy who's currently running it to finish successfully, but you never know. |
|
|
|
|
I manually ran that workunit here and inserted the result. Don't read anything into the ratio between CPU and elapsed time because CPU = elapsed * 4 since I used four cores and all I had was the elapsed time from the LLR outcome string. I did this because I don't really expect the guy who's currently running it to finish successfully, but you never know.
Just curiosity: can an average user run a test using 4 cores by BOINC? How could I enable multithreading if I desire to do it? App_config.xml... cmdline tag? |
|
|
RafaelVolunteer tester
Send message
Joined: 22 Oct 14
Posts: 911
ID: 370496
Credit: 550,234,793
RAC: 442,833
|
I manually ran that workunit here and inserted the result. Don't read anything into the ratio between CPU and elapsed time because CPU = elapsed * 4 since I used four cores and all I had was the elapsed time from the LLR outcome string. I did this because I don't really expect the guy who's currently running it to finish successfully, but you never know.
Just curiosity: can an average user run a test using 4 cores by BOINC? How could I enable multithreading if I desire to do it? App_config.xml... cmdline tag?
http://www.primegrid.com/forum_thread.php?id=7348&nowrap=true#106105 |
|
|
|
|
http://www.primegrid.com/forum_thread.php?id=7348&nowrap=true#106105
Thank you. That was exactly what I was thinking about. This method will kill the competition on LLR tests, but it will cost efficiency too.
/OT |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 433,172,407
RAC: 1,010,059
|
http://www.primegrid.com/forum_thread.php?id=7348&nowrap=true#106105
Thank you. That was exactly what I was thinking about. This method will kill the competition on LLR tests, but it will cost efficiency too.
/OT
Surprisingly, you may find it increases efficiency.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
And now it is: two valid, three invalid, and one in-progress. Stay tuned.
____________
I'm counting for science,
Points just make me sick. |
|
|
JimBHonorary cruncher
Send message
Joined: 4 Aug 11
Posts: 920
ID: 107307
Credit: 989,246,873
RAC: 201,706
|
|
No need to stay tuned. Mike and I both ran that candidate independently and got the same exact result. Since mine finished first, I inserted my residue into BOINC. The job in progress is currently 56% done and he'll either get credit for agreeing with the canonical result or he won't. The odds of that last job randomly agreeing with one of the other residues is 1 in 18446744073709551616 or so. |
|
|
|
|
|
And I'll keep you updated anyway. Unless I'd die before.
____________
I'm counting for science,
Points just make me sick. |
|
|
|
|
|
He did it! Now it's a tie (no, probably, it's not): three valids and three invalids. Hooray for UID193455 everyone!
____________
I'm counting for science,
Points just make me sick. |
|
|
