| Author |
Message |
|---|
 Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
                           
|
|
We've increased the deadline for SR5 tasks to 6 days. With larger numbers come longer processing times. Since it also takes up to 4 times the normal time to process a number when a prime is found, we feel this increase in the deadline is warranted.
Since there's still several hundred SR5 workunits in the pipeline, you may still get some SR5 tasks with 3 day deadlines, but all new workunits will have a 6 day deadline.
Note: Some older CPUs such as an AMD Sempron may run SR5 tasks up to four times slower than they run other LLR tasks. Therefore, if you have a really old computer (like my 32-bit Sempron) and are lucky enough to find a Riesel prime, you won't be able to complete the processing within the deadline. If your computer can't complete an SR5 task in 36 hours (1/4th of the deadline), it won't be able to make the deadline if it finds a Riesel (-1) prime. Sierpinski (+1) primes "only" take twice as long.
When deciding whether to process SR5, please keep this extra processing time for primes in mind. If you have a very slow processor and miss the deadline, you may miss out on getting credit for discovering the prime or being the double checker.
____________
My lucky number is 75898524288+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
|
While reviewing the most recent prime, it looks like I need to revise my statement about primes taking four times as long to calculate.
The most recent prime took almost ten times as long to calculate.
Therefore, so that people have enough time to process these numbers, I'm increasing the deadline (again) to 15 days.
So, if you notice that your SR5 job is taking an unusually long time, or that your wingman took a really long time, it's probably because you've just found the next SR5 prime!
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Does this ten times longer test is on AVX or non-avx cpu?
When that prime is reported I will test it , and see run times...
____________
92*10^1439761-1 REPDIGIT PRIME :) :) :)
314187728^131072+1 GENERALIZED FERMAT
31*332^367560+1 CRUS PRIME
Proud member of team Aggie The Pew. Go Aggie! |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
Does this ten times longer test is on AVX or non-avx cpu?
When that prime is reported I will test it , and see run times...
Both.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Task is past the 2 day mark. Hope it's a prime and not some kind of error. The progress % hasn't moved in awhile |
|
|
JimBHonorary cruncher
Send message
Joined: 4 Aug 11
Posts: 916
ID: 107307
Credit: 974,514,092
RAC: 0
         
|
Task is past the 2 day mark. Hope it's a prime and not some kind of error. The progress % hasn't moved in awhile
My impression is that the task isn't running at all. It was 72% done on 1 December. Exactly 24 hours later it was still at 72% done. That's not how a prime would behave. A prime would restart at zero percent done after it hit 100%. |
|
|
|
|
|
Units normally take 2hr 15 to process on one machine. I have one now at 1d 4:27:45. It's showing 100% so not sure if it's stuck or doing something.
I was at about 98% 4 hours ago. |
|
|
|
|
Units normally take 2hr 15 to process on one machine. I have one now at 1d 4:27:45. It's showing 100% so not sure if it's stuck or doing something.
I was at about 98% 4 hours ago.
Let it finish. If you're lucky, you might have just found a prime number! |
|
|
|
|
|
Will LLR say something in stderr.txt if it starts the primality test?
____________
SHSID Electronics Group
SHSIDElectronicsGroup@outlook.com
GFN-14: 50103906^16384+1
Proth "SoB": 44243*2^440969+1
 |
|
|
|
|
|
Now that SR5 candidates have become huge, would it be possible and make sense to change the way LLR is invoked so that it does not do anything other than the PRP test in any case? It would then be necessary for the server to run the deterministic primality test.
Just a thought; it might be a bad idea, or impossible.
I think we have other subprojects that work like that. Genefer?
/JeppeSN |
|
|
|
|
Units normally take 2hr 15 to process on one machine. I have one now at 1d 4:27:45. It's showing 100% so not sure if it's stuck or doing something.
I was at about 98% 4 hours ago.
It's stuck. I've had this happen to me. If it were prime the percentage is supposed to start over for the second test. Reboot the computer and it will unstick the task.
____________
Eating more cheese on Thursdays. |
|
|
|
|
It's stuck. I've had this happen to me. If it were prime the percentage is supposed to start over for the second test. Reboot the computer and it will unstick the task.
Rebooting sounds like a good idea. You will not lose the task by rebooting the computer. /JeppeSN |
|
|
|
|
|
Thanks all done so and it's gone |
|
|
|
|
Will LLR say something in stderr.txt if it starts the primality test?
Doesn't look like it. If you look at the recent SR5 tasks where the number came out as prime, there's nothing in the stderr output indicating when a primality test began. |
|
|
|
|
|
Oh, just asking ;)
Not that I may get a SR5 prime someday :)
____________
SHSID Electronics Group
SHSIDElectronicsGroup@outlook.com
GFN-14: 50103906^16384+1
Proth "SoB": 44243*2^440969+1
|
|
|
JimBHonorary cruncher
Send message
Joined: 4 Aug 11
Posts: 916
ID: 107307
Credit: 974,514,092
RAC: 0
|
|
This is what they look like if you run them manually:
jim@ps2:~/test$ ./c -d -t4 -q"146264*5^2953282-1"
Base prime factor(s) taken : 5
Starting N+1 prime test of 146264*5^2953282-1
Using FMA3 FFT length 720K, Pass1=320, Pass2=2304, clm=4, 4 threads, a = 3
146264*5^2953282-1 may be prime. Starting Lucas sequence...
Using FMA3 FFT length 720K, Pass1=320, Pass2=2304, clm=4, 4 threads, P = 4
146264*5^2953282-1 may be prime, trying to compute gcd's
U((N+1)/5) is coprime to N!
146264*5^2953282-1 is prime! (2064261 decimal digits, P = 4) Time : 23182.986 sec. |
|
|
|
|
|
I think it's the -d switch that does it right? Should I add it into my app config?
____________
SHSID Electronics Group
SHSIDElectronicsGroup@outlook.com
GFN-14: 50103906^16384+1
Proth "SoB": 44243*2^440969+1
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
|
I think it's the -d switch that does it right? Should I add it into my app config?
There is nothing you can do that will cause that output to be visible to you. It's eaten by the wrapper. At worst, you'll confuse the wrapper and/or the validator, and the task will fail.
At best, it will have no effect because the wrapper is already invoking LLR with the -d parameter.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Ok, I never knew the wrapper did that.
Good to know! Thanks!
____________
SHSID Electronics Group
SHSIDElectronicsGroup@outlook.com
GFN-14: 50103906^16384+1
Proth "SoB": 44243*2^440969+1
|
|
|
|
|
|
I browsed through the documentation of LLR. It seems there is an option ForcePRP=1 with which LLR will do only the PRP test.
If PrimeGrid started to use that option (and if that option actually works the way I claim it does), then primes would not take longer than composites anymore.
The burden of checking if the PRP was really a prime, would then be moved to the PrimeGrid server.
I do not know if this would be a good idea; I will leave it as an open question.
Maybe the number of false positives would be too high if each client did only one test?
/JeppeSN |
|
|
|
|
|
I believe most of us are here to find primes, not just probable primes. Also I feel that we are donating our computers to do the processing whereas doing only half the work and then sloughing it off onto the PG servers defeats that purpose.
____________
Werinbert is not prime... or PRPnet keeps telling me so.
Badge score: 1x1 + 12x3 + 1x4 + 1x5 + 1x6 + 2x7 + 1x8 + 1x10 = 84 |
|
|
JimBHonorary cruncher
Send message
Joined: 4 Aug 11
Posts: 916
ID: 107307
Credit: 974,514,092
RAC: 0
|
|
I just realized we hadn't said anything about credit. Because we know an SR5 prime in the form k*5^n+1 takes twice as long, we give twice the BOINC credit. For an SR5 prime in the form k*5^n-1 we give 5x the normal credit. So nobody is missing out on credit for discovering a conjecture prime (if that would even bother anyone in the first place). |
|
|
