Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Sierpinski/Riesel Base 5 Problem :
SR5 prime found or stuck task?
Author |
Message |
|
For the past 36 hours, my wingman and I have been pounding away on this SR5 task on my 3930k:
http://www.primegrid.com/workunit.php?wuid=386276075
From the sticky thread, I know finding a prime is supposed to take 10x as long to crunch as a regular workunit, which take around 2:15 on my computer (6 threads active). This particular one was at about 44% at the 2 hour mark, and the percentage per hour completion rate steadily decreased and seems to be asymptotic with 0. To speed the crunching and get that coveted #1 spot (since I am against an Ivy i5) I suspended the other tasks and have just the 1 core running on this guy. Yesterday afternoon at 25 hours it was 98%, this morning at 36 hours when I left for work it was at 99.994%. Based on the other SR5 prime times on a 3930k, that seems a bit long, yes?
So, I'm curious, considering my wingman is still working on this as well, is it likely that we are close to finding the next SR5 or is something off here?
____________
Eating more cheese on Thursdays. | |
|
JimB Honorary cruncher Send message
Joined: 4 Aug 11 Posts: 916 ID: 107307 Credit: 974,494,172 RAC: 310
                    
|
It's certainly possible that it's prime, but the only way I could answer this is to test that candidate myself. I don't know how the boinc client displays progress on an SR5 prime as I've never had one myself. I've triplechecked quite a few, but we generally do so manually using the pfgw program.
Your wingman could be running any number of projects simultaneously, so you can't really conclude anything from how long he/she is taking to return that candidate. | |
|
Michael Goetz Volunteer moderator Project administrator
 Send message
Joined: 21 Jan 10 Posts: 13527 ID: 53948 Credit: 245,568,490 RAC: 284,022
                          
|
I've started checking it on this machine. I'll let you know in about 90 minutes.
____________
My lucky number is 75898524288+1 | |
|
|
Your wingman could be running any number of projects simultaneously, so you can't really conclude anything from how long he/she is taking to return that candidate.
True. I checked their FreeDC computer stats and Primegrid tasks list and it has him/her at 90% Primegrid recently. Best I can figure, it means that they seem to be spending a lot of time on the same task, but that could be just some bad table-top math :)
Michael-I look forward to your results! I hope the 44+ hours I'm currently at won't be wasted. :)
____________
Eating more cheese on Thursdays. | |
|
Michael Goetz Volunteer moderator Project administrator
 Send message
Joined: 21 Jan 10 Posts: 13527 ID: 53948 Credit: 245,568,490 RAC: 284,022
                          
|
Result sent via PM.
In the future, it's best to ask these questions via PM rather than publicly. It's possible (although unlikely) that an unscrupulous person could figure out what number you're testing and attempt to report the prime as their discovery to the top 5000 prime list before you and your wingman complete the testing.
In general, if a task appears to be "stuck", the best approach is:
1) In the BOINC manager, "Suspend Activity" to stop all running tasks.
2) Reboot your computer.
3) In the BOINC manager, change back to "Run according to preferences."
In most cases, that should unstick the task.
Although I've never personally seen it, on an SGS or SR5 prime, what you should see if a prime is found is a normal progression from 0 to 99% followed by the percentage resetting back to 0, and then progressing a second time up to 100%. In the case of an SR5 test on a Reisel number (i.e., c = -1), this will occur about 5 times.
____________
My lucky number is 75898524288+1 | |
|
axnVolunteer developer Send message
Joined: 29 Dec 07 Posts: 285 ID: 16874 Credit: 28,027,106 RAC: 0
            
|
Is it my imagination or did the OP abort his task ?!
EDIT:- That means it is not prime? :-( | |
|
serge  Send message
Joined: 21 Jun 12 Posts: 112 ID: 144858 Credit: 242,069,945 RAC: 144,198
                   
|
Not this one, but another one is. (by Brown. Congrats!)
____________
My lucky number is Phi(4, 2^2396029-1)/2.
| |
|
Scott Brown Volunteer moderator Project administrator Volunteer tester Project scientist
 Send message
Joined: 17 Oct 05 Posts: 2183 ID: 1178 Credit: 9,122,702,041 RAC: 12,810,911
                                      
|
Not this one, but another one is. (by Brown. Congrats!)
Yep. :)
Official announcement and news post should be up soon. That's 4 so far this year...anybody think we can make it a dozen before 2014 is over?
| |
|
|
Another "K" bites the dust.
____________
My lucky numbers are 121*2^4553899-1 and 3756801695685*2^666669±1 | |
|
Michael Goetz Volunteer moderator Project administrator
 Send message
Joined: 21 Jan 10 Posts: 13527 ID: 53948 Credit: 245,568,490 RAC: 284,022
                          
|
Another "K" bites the dust.
Two more k's bit the dust today. More information will be released once wingmen, reporting, verification, etc. is done.
____________
My lucky number is 75898524288+1 | |
|
axnVolunteer developer Send message
Joined: 29 Dec 07 Posts: 285 ID: 16874 Credit: 28,027,106 RAC: 0
            
|
One of the primes has been reported to Top-5000: http://primes.utm.edu/primes/page.php?id=117597
Looking at the status page, it would seem that the other prime is for 207394*5^n-1 | |
|
axnVolunteer developer Send message
Joined: 29 Dec 07 Posts: 285 ID: 16874 Credit: 28,027,106 RAC: 0
            
|
For the second one: Are we waiting on doublechecker to finish up (unlikely), or on the primary finder to report it (19 days)?
Can we at least announce the first one? | |
|
Honza Volunteer moderator Volunteer tester Project scientist Send message
Joined: 15 Aug 05 Posts: 1893 ID: 352 Credit: 3,285,613,344 RAC: 5,208,802
                              
|
For the second one: Are we waiting on doublechecker to finish up (unlikely), or on the primary finder to report it (19 days)?
We are still waiting for a wingman to doublecheck.
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186 | |
|
Scott Brown Volunteer moderator Project administrator Volunteer tester Project scientist
 Send message
Joined: 17 Oct 05 Posts: 2183 ID: 1178 Credit: 9,122,702,041 RAC: 12,810,911
                                      
|
Can we at least announce the first one?
That's my fault. Real life has been unusually hectic this week, but I should have it done tonight after a grueling 3.5 hour meeting to end the work week. :(
| |
|
Michael Goetz Volunteer moderator Project administrator
 Send message
Joined: 21 Jan 10 Posts: 13527 ID: 53948 Credit: 245,568,490 RAC: 284,022
                          
|
For the second one: Are we waiting on doublechecker to finish up (unlikely), or on the primary finder to report it (19 days)?
We are still waiting for a wingman to doublecheck.
Due in large part to this prime, as well as concern about what could happen when we find a PSP or SoB prime (which can take months to validate), we've modified our procedures somewhat.
Up until now, the procedure for handling discoveries of large primes (i.e., mega primes) on BOINC went like this:
1) Someone discovers a prime.
2) We see that ONE person discovered a prime, but it's still waiting for the wingman. Unless we suspect it's a false positive, we'll start verifying the prime on our own computers. Although we typically say "we're verifying it on the server...", we really mean we're verifying it on our personal computers because they're MUCH faster than the server. (The server isn't AVX.)
3) When the wingman comes in, we notify the prime finder. If he has authorized us to report on his behalf, we proceed to step 4. Otherwise, we send him an email and he either authorizes us to report for him, or he reports it himself. If he doesn't respond in 19 days, we repeat the process with the wingman.
4) Either the user reports the prime to the top 5000 website, or we report it for him or her.
5) We update the appropriate PrimeGrid webpages and publish an announcement about the prime.
6) When the prime is verified on the top 5000 website, the PrimeGrid server makes the prime visible in the prime list.
The problem is that if you have a wingman who uses up the entire deadline, and then doesn't return a result, you just wait and wait and wait. For the really large numbers, that can take months if you're unlucky.
This is the new procedure:
1) Someone discovers a prime.
2) We see that ONE person discovered a prime, but it's still waiting for the wingman. Unless we suspect it's a false positive, we'll start verifying the prime on our own computers. Although we typically say "we're verifying it on the server...", we really mean we're verifying it on our personal computers because they're MUCH faster than the server. (The server isn't AVX.)
3) NEW: Once we've verified that the number is indeed prime (using both LLR and PFGW), if the wingman hasn't responded yet, we modify the workunit so it only requires one task to verify. The prime number then gets validated immediately without the need for a wingman. This permits us to proceed with reporting the prime. The wingman's task is allowed to continue and he will get BOINC credit and be recognized as the double checker if he correctly returns the result.
4) Without waiting for the wingman we notify the prime finder. If he has authorized us to report on his behalf, we proceed to step 5. Otherwise, we send him an email and he either authorizes us to report for him, or he reports it himself.
5) Either the user reports the prime to the top 5000 website, or we report it for him or her.
6) We update the appropriate PrimeGrid webpages and publish an announcement about the prime.
7) When the prime is verified on the top 5000 website, the PrimeGrid server makes the prime visible in the prime list.
What the change means is that there won't be a potentially endless series of bad hosts failing to return the double check task. The double check that is in progress is allowed to finish, and if he returns the result he'll get credit for the double check. But if he doesn't return the result, no additional tasks will be sent out for the purpose of double checking.
____________
My lucky number is 75898524288+1 | |
|
axnVolunteer developer Send message
Joined: 29 Dec 07 Posts: 285 ID: 16874 Credit: 28,027,106 RAC: 0
            
|
I would go one further. For SR5, if LLR permits it*, only do the initial part of test on the client. Skip the proving part, and report the results straightaway. The admins will be doing the rigorous proof offline, so nothing's lost. The total time to report the prime will come down. This will also obviate this thread (long running task = stuck!)
* Whether LLR can be configured for only the first part: There are three settings that I can find in the documentation
ForcePRP=1 : Do only a PRP test, even if a deterministic one is possible.
LucasPRPtest=1 : For a PRP test, use only the Lucas+Frobenius algorithm.
FermatPRPtest=1 : For a PRP test, use only the Fermat SPRP algorithm.
| |
|
Michael Goetz Volunteer moderator Project administrator
 Send message
Joined: 21 Jan 10 Posts: 13527 ID: 53948 Credit: 245,568,490 RAC: 284,022
                          
|
Here's the second SR5 prime from April 9th: http://primes.utm.edu/primes/page.php?id=117731
Congratulations Honza!
207394*5^1612573-1
We can scratch k=207394 off the Riesel base 5 list.
____________
My lucky number is 75898524288+1 | |
|
Dave  Send message
Joined: 13 Feb 12 Posts: 2842 ID: 130544 Credit: 990,268,704 RAC: 1,311,864
                      
|
Yet another SR5 k has been obliterated! | |
|
Michael Goetz Volunteer moderator Project administrator
 Send message
Joined: 21 Jan 10 Posts: 13527 ID: 53948 Credit: 245,568,490 RAC: 284,022
                          
|
This one was found on the 25th: http://primes.utm.edu/primes/page.php?id=117737
Congratulations Scott!
____________
My lucky number is 75898524288+1 | |
|
Honza Volunteer moderator Volunteer tester Project scientist Send message
Joined: 15 Aug 05 Posts: 1893 ID: 352 Credit: 3,285,613,344 RAC: 5,208,802
                              
|
Congrats Scott.
Looks like we have made TOP5000 verification process busy for a while :-)
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186 | |
|
Message boards :
Sierpinski/Riesel Base 5 Problem :
SR5 prime found or stuck task? |