## Other

drummers-lowrise

Message boards : Seventeen or Bust : Crunching k=10223 and k=67607

 Subscribe SortOldest firstNewest firstHighest rated posts first
Author Message
Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 82989 - Posted: 28 Jan 2015 | 16:33:26 UTC

After some discussion with Louie Helm at Seventeen or Bust, we are going to change the way work is divided between SoB and PG. After a transition period, PrimeGrid will be crunching only k=10223 and k=67607, while SoB will crunch only the other 4 k's. Each project will be entirely responsible for its own k's.

Why? Well, if we never find another SoB prime, things will work fine the way are now. But if we take a more optimistic view and assume we WILL find a prime, there's a problem: We'll have wasted an immense amount of crunching. This change will correct that problem.

Currently, PrimeGrid is crunching in the n=27M range, and SoB is crunching in the 29M range. If PrimeGrid finds a prime in the 27M range, every test that SoB did in the 28M and 29M ranges in that k will have been wasted effort! Likewise, once we finish the 27M tests and move on to 31M tests, if SoB finds a prime, everything PrimeGrid will have tested at 31M in that k will have been wasted.

By dividing the work by k we avoid this problem.

One thing that doesn't change: if PrimeGrid finds a prime, both PG and SoB share the project credit for the discovery.

The current plan is for us to complete the current 27M and upcoming 31M tests in all the k's and then only crunch 10223 and 67607. At some point we will probably do some double checking of earlier SoB work that was not yet double checked. Details will become available later.

(For those that are curious and more attuned to the details, the residues that SoB uses are not the same as what PrimeGrid normally uses. This makes double checks more complicated. However, our LLR program can do the same test and produce the same residue as SoB's software if we use the correct command line arguments. However, SoB's older software has completely incompatible residues, so hopefully there aren't a lot of those that need double checking.)
____________
My lucky number is 75898524288+1

Honza
Volunteer moderator
Volunteer tester
Project scientist

Joined: 15 Aug 05
Posts: 1893
ID: 352
Credit: 3,141,484,578
RAC: 2,037,366

Message 83007 - Posted: 29 Jan 2015 | 9:26:20 UTC

Agree, it seems to be a more effective way, also in term of no longer reserving ranges etc.

Is it expected (or desirable) that both PG and SoB maintaine about the same progress for each k?

(for sure, once a k is eliminated, we might need to reconsider splitting other k's).

Is sieving done up to 50M?
If there is need for more, I assume PG will do more sieving and provide remaining candidates to SoB.
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 83010 - Posted: 29 Jan 2015 | 10:30:44 UTC - in response to Message 83007.

Is it expected (or desirable) that both PG and SoB maintaine about the same progress for each k?

Desirable? Not sure. We've punched certain k's higher for some reason in the past in conjectures. In the end, it doesn't matter much what order you search the k's because to solve the conjecture you need to do all those tests anyway. In a sense, you could treat each k as an individual project.

The only thing binding them together is sieving. We're done with sieving up to 50M, so that's not a factor anymore.

It's almost certain that PG and SoB will not advance at the same speed, so one of us will definitely be at a higher n,

(for sure, once a k is eliminated, we might need to reconsider splitting other k's).

That's the reason we have 2 k's instead of 1 k. When we find a prime in one of the 2 k's, we still have another k to crunch while we figure out which remaining k (if any) to transfer from SoB to PG.

Is sieving done up to 50M?

Yes. We're sieving ESP up to 50M too, but SoB and PSP are completely sieved to 50M.

____________
My lucky number is 75898524288+1

Pooh Bear 27

Joined: 10 May 09
Posts: 709
ID: 39821
Credit: 407,730,319
RAC: 8,998

Message 83052 - Posted: 30 Jan 2015 | 2:50:25 UTC

What does this mean in time and deadline?

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 236,922,854
RAC: 160,213

Message 83062 - Posted: 30 Jan 2015 | 11:52:12 UTC - in response to Message 83052.

What does this mean in time and deadline?

Immediately, nothing. I expect some double check tasks will be a lot shorter and likely have much shorter deadlines, but details are not available yet.
____________
My lucky number is 75898524288+1

JimB
Honorary cruncher

Joined: 4 Aug 11
Posts: 916
ID: 107307
Credit: 974,494,092
RAC: 3,785

Message 86243 - Posted: 15 Jun 2015 | 19:42:44 UTC