Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Sieving :
Suspending Factorial and Primorial sieving November 1
| Author |
Message |
|---|
JimB Honorary cruncher
Send message
Joined: 4 Aug 11
Posts: 920
ID: 107307
Credit: 989,270,184
RAC: 150,909
                     
|
|
I've been looking into how well sieved we are for factorial and primorial. While at first glance the optimal sieving point appears to be p=376T for factorial and p=133T for primorial, that's only if we're going to test up to the maximum candidate size. For where we are right now, we're oversieved. In the time it takes to remove a candidate by CPU sieving, one can test 3.27 factorial candidates or 1.64 primorial candidates. By definition, when it takes longer to sieve than to test, it's time to stop sieving.
Being on PRPNet, primorial and factorial testing don't progress very quickly. Since LLR won't work on those types of candidates, PFGW is used. We don't currently use PFGW for BOINC testing and there are no plans to do so. So they're going to stay on PRPNet. Sieving for a candidate size that we might reach in 25+ years doesn't seem to be a good use of processing power right now.
Effective November 1st, no new sieving reservations will be accepted for either factorial or primorial sieving. Sieving will be suspended until such time as it's needed again. Obviously that'll be sooner for primorial as the testing/sieving ratio is smaller.
Feel free to discuss this, but from my point of view it's hard to defend more sieving at this point in time. | |
|
 RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1138
ID: 120786
Credit: 268,668,824
RAC: 0
                 
|
|
Good work with the new Manual Sieving system JimB. I'll have to try it soon.
My argument against removing Factorial Prime Sieving is it's one of the few apps for AMD GPUs.
I found the bug that prevents sieve input file from ever being able to work with fsievecl:
http://www.primegrid.com/forum_thread.php?id=1163&nowrap=true#80600 | |
|
RafaelVolunteer tester
Send message
Joined: 22 Oct 14
Posts: 912
ID: 370496
Credit: 552,485,027
RAC: 466,454
                  
|
|
Uh... if we are already oversieved, why don't we just stop right now? Why wait till Nov? | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 435,681,308
RAC: 871,866
               
|
Uh... if we are already oversieved, why don't we just stop right now? Why wait till Nov?
From a purely mathematical standpoint, there's no reason. But when we shut down a project, try to give participants a bit of advance notice so they have an opportunity to do some last-minute tasks. That's more important when we're shutting down a project with its own badge, but we just want to give everyone a chance to get some work in before it's shut down.
Also, the decision about where to stop sieving is very much a judgement call, so there's no precise spot where you suddenly go from sieving being useful to sieving being useless. It's very much a continuum rather than a specific point.
____________
My lucky number is 75898524288+1 | |
|
HAmstyVolunteer tester
Send message
Joined: 26 Dec 08
Posts: 132
ID: 33421
Credit: 12,510,712
RAC: 0
                
|
|
fpsieve is the last linux cpu target for offline use on primegrid. Apart from that, no hard feelings on my side.
____________
 | |
|
JimBHonorary cruncher
Send message
Joined: 4 Aug 11
Posts: 920
ID: 107307
Credit: 989,270,184
RAC: 150,909
|
|
Roger wrote: I found the bug that prevents sieve input file from ever being able to work with fsievecl:
http://www.primegrid.com/forum_thread.php?id=1163&nowrap=true#80600
Having fsievecl take a sieve input file is useless. All it does, as far as I know, is to suppress output of factors where the candidate is not in the sieve. It will not speed up the process at all. I'd prefer to have all those factors output, it's not like these are huge files. Every time someone hands in an empty factor file, I have to rerun it here myself to ensure they did it properly. I like having all those extra factors. | |
|
|
|
|
Well, we all knew that this moment will come. If you say we reach a point when we oversieveing, that is no sense into waste more resources and it's time to shut it down. Thanks for heads up though.
So guys, if this are last weeks of it, maybe we push this numbers up?
I call 10 000G for primorial. :D | |
|
|
|
For where we are right now, we're oversieved. In the time it takes to remove a candidate by CPU sieving, one can test 3.27 factorial candidates or 1.64 primorial candidates. By definition, when it takes longer to sieve than to test, it's time to stop sieving.
On Dual E5-2620 (running 12 threads), times are per 1 thread:
fpsieve - 1factor/24hrs
pfgw64 - 1test/20hrs | |
|
Message boards :
Sieving :
Suspending Factorial and Primorial sieving November 1 |
