Search for k's at n>1M

[em99010pepe]

Gary,

Do you have sieved files from k's beyond 1M? Just curiosity.

Carlos

[gd_barnes]

Originally Posted by em99010pepe

Gary,

Do you have sieved files from k's beyond 1M? Just curiosity.

Carlos

No. Bruce is currently sieving 600K-1M for k=400-1001. That's as close as we get.

If you are interested in something similar to RPS's 6th drive with several k's at once searched for n>600K, we can work something out on the individual-k drive for the remaining unreserved k's.


Gary

[em99010pepe]

Let's organize a search beyond 1M for one k. Let's find a 1,000,000 digit prime, shall we? I propose a lower weight k.

[henryzz]

Originally Posted by em99010pepe

Let's organize a search beyond 1M for one k. Let's find a 1,000,000 digit prime, shall we? I propose a lower weight k.

isnt that going against the whole plan of nlpb

[em99010pepe]

Originally Posted by henryzz

isnt that going against the whole plan of nlpb

Please post here the plans of NPLB project.

[henryzz]

Originally Posted by em99010pepe

Please post here the plans of NPLB project.

to take large groups of ks to steadily decreasing n values

i could understand taking a very low k to >1M if the ks surrounding it are >1M but taking say 5000 to 1M is how RPS works not NLPB

[em99010pepe]

The goal of NPLB project is to take 300 < k < 1001 tested up to n = 1000000 until the end of 2012.

But you can pick one k and take it beyond 1M. We already have a member doing it but I was thinking in making a group effort with sieving and LLRing.

Carlos

[Flatlander]

Maybe sieve a bunch of specific ks?

[gd_barnes]

Although I'd prefer to not make a specific drive out of it, doing a coordinated sieving effort for and then searching k's for n>1M is fine.

If you're going to do low-weights, I would suggest picking them from the k=300-400 range.

Another thought is to pick some extreme low weights for k>1001. I'm just about done with a "side effort" to find primes for k's that show zero primes on the rieselprime.org pages for k=600K-1.4M. I'm searching them to n=100K and if no prime found taking them to n=300K. I did a large majority of it way back but am finishing it up now.

With this go around, I've searched 17 low weight k's that have no primes shown there and have found just one k remaining that has no prime up to n=300K. Previously I searched a whole bunch of them (don't remember the #) for k=600K-1M up to n=300K, and found 9 k's with no primes. For everyone's reference, here are 10 k's for k=600K-1.4M with no primes up to n=300K:

612509
671413
685183
686711
700057
780427
844559
963643
981493
1049917*

*The last one was supposedly searched by someone else already to n=1M but I've learned not to trust these ranges. I've found primes on many of them that had been shown as searched to limits higher than the primes that I found.

In the next day or two, I'll be reporting all the primes that I found for the effort and after Karsten gets the site updated, the above k's should be the only ones remaining shown on the site for k=600K-1.4M with no primes to n=300K.

Whether it be for k=300-400 (preferred) or for k=600K-1.4M, this provides some food for thought if you guys want to coordinate some sieving on low-weight k's. All except 2 of the 10 k's that I showed above have weights < 100.

One thing about the larger k's is that they will take longer to test, which will be quite significant at n>1M so lower k's would certainly be preferrable.

One final suggestion: Try a HEAVY-weight for a k=300-400 that has already been completed to n=1M or that is ABOUT to become completed. (My k=309 would be a decent choice.) I mention heavy-weight because you could get a lot of tests in at the lower n>1M ranges. Searching, perhaps, 5000-10000 candidates in the n=1M-1.2M range gives a much better chance at prime than the same # of searches for n=1M-2M.


Gary

[gd_barnes]

Originally Posted by em99010pepe

Please post here the plans of NPLB project.

See http://www.mersenneforum.org/showthread.php?t=10618


#2 is done. #4 has been opened up already. It's pretty much set in stone that we'll do k=400-1001 for n=600K-1M in 3 separate drives of k=400-600, 600-800, & 800-1001 as people seemed to like that idea. As always, I'm open to suggestions though.

k=1005-2000 for n=50K-500K hasn't been fully decided yet. I know we'll search all of the k's in tandem at the lower n-ranges as we continue to sieve. At n=50K, the ranges will fly even with ~500 k's. I'm thinking we'll open up the top-5000 range for n=350K-500K after we have sieved far enough but this entire effort is still open to team discussion on how to progress with it.

One reason that I'm not specifically promoting n>1M searches is that the above is a huge amount of work already. But I realize that people have different tolerances for what they like to search so it's fine if you guys want to take up a coordinated effort for them.


Gary

[henryzz]

one thing i would say dont chose k<300 as RPS will go berserk