LLRnet servers for NPLB

[gd_barnes]

It's nice to see the positive response to looking for some smaller top-5000 primes. I think that settles it that we'll put a push on sieving the k=1005-2000 range after the 1st drive is done.

Ian, if you don't mind, go right ahead and continue sieving above P=1T. After the 1st drive is done, I'll put at least 2 cores on it. Within a few days after the drive is done, I'll start a thread for a public sieving effort to expidite it even more. With that, I'm confident that we can get it up to P=4T-6T before the end of Jan.

I do want to mention to everyone that the top-5000 part of the effort will be ~35-40% double-check up to n=500K but is necessary due to the willy-nilly nature of how the n-ranges were searched for this k-range. But as fast as the tests will be at the lower n-level, we'll still find a lot of new primes. The lower, non-top-5000 part, i.e. n=50K-350K, will only be ~10% double-check from what I can tell.

Max, yes, we'll start testing n=50K-350K while we're still sieving the higher n-range. Actually, the lower part is way past optimal sieve depth now but it made sense to just leave it in there. When we start primality testing the lower part, we'll just break off n=50K-100K to start with and I'll post a new sieve file with it removed...no use LLRing and sieving the same range.


Gary

[gd_barnes]

Originally Posted by mdettweiler

Meanwhile, I've marked port 8000 as "for future use" rather than "NPLB 7th Drive" on the status page, to reflect its new purpose.

Max

Let's see, we'll number the drives here:

k=400-600 for n>600K; 5th drive
k=600-800 for n>600K; 6th drive
k=800-1001 for n>600K; 7th drive
k=1005-2000 for n=50K-350K; 8th drive
k=1005-2000 for n=350K-500K; 9th drive

Anyone have any better ideas?

Assuming this makes sense and people agree with the numbering scheme, you can mark port 8000 for the 9th drive if you want.


Gary

[mdettweiler]

Originally Posted by gd_barnes

Let's see, we'll number the drives here:

k=400-600 for n>600K; 5th drive
k=600-800 for n>600K; 6th drive
k=800-1001 for n>600K; 7th drive
k=1005-2000 for n=50K-350K; 8th drive
k=1005-2000 for n=350K-500K; 9th drive

Anyone have any better ideas?

Assuming this makes sense and people agree with the numbering scheme, you can mark port 8000 for the 9th drive if you want.


Gary

Okay, that looks good. One suggestion, though: how about we combine the 8th and 9th drives into one drive (8th)? Sort of like what we did for the 2nd Drive, with two n-ranges listed separately in one drive thread? That might cut down on the number of required sticky threads, which is already going to be enormous.

Max

[MyDogBuster]

Ian, if you don't mind, go right ahead and continue sieving above P=1T.

Sieving starting at 1T. Of course this means I took 2 cores off G4000. I'll gain them back with my finishing the manual ranges.

[gd_barnes]

Originally Posted by mdettweiler

Okay, that looks good. One suggestion, though: how about we combine the 8th and 9th drives into one drive (8th)? Sort of like what we did for the 2nd Drive, with two n-ranges listed separately in one drive thread? That might cut down on the number of required sticky threads, which is already going to be enormous.

Max

Hum. I confused. When did we list two n-ranges in one drive thread? The 1st drive is k=400-1001 for n=332K-600K, the 2nd drive was k=400-1001 for n=260K-332K, and the 3rd drive was k=300-400 for n=260K-600K. We searched the 3rd drive almost straight up from n=260K. Although I see that we did allow n>=320K reservations initially, Beyond came in and quickly reserved a huge range and finished off the lower range. I think we might have had 2 n-ranges for a bit on the 2nd drive, one for n=320K-332K to fill in the n=20K reporting range at prime search and the other for n=260K-320K but that was small and it was fairly quickly down to one n-range.

We're talking a huge n-range gap here for n=50K-350K and 350K-500K. They should probably be 2 drives for the same reason that we had 2 drives for k=400-1001.

That said, I agree the stickies will be huge. When we start adding drives, let's discuss which threads can be unstickied.

Originally Posted by MyDogBuster

Sieving starting at 1T. Of course this means I took 2 cores off G4000. I'll gain them back with my finishing the manual ranges.

No problem. We're in good shape on port 4000. When you finish your manual ranges, I'll move my quad back to port 400.


Gary

[henryzz]

to help solve the stickies problem would it be worth having one stickied thread that has links to several threads that would be sticked otherwise

[mdettweiler]

Originally Posted by gd_barnes

Hum. I confused. When did we list two n-ranges in one drive thread? The 1st drive is k=400-1001 for n=332K-600K, the 2nd drive was k=400-1001 for n=260K-332K, and the 3rd drive was k=300-400 for n=260K-600K. We searched the 3rd drive almost straight up from n=260K. Although I see that we did allow n>=320K reservations initially, Beyond came in and quickly reserved a huge range and finished off the lower range. I think we might have had 2 n-ranges for a bit on the 2nd drive, one for n=320K-332K to fill in the n=20K reporting range at prime search and the other for n=260K-320K but that was small and it was fairly quickly down to one n-range.

We're talking a huge n-range gap here for n=50K-350K and 350K-500K. They should probably be 2 drives for the same reason that we had 2 drives for k=400-1001.

That said, I agree the stickies will be huge. When we start adding drives, let's discuss which threads can be unstickied.

Hmm...I see what you mean. Yeah, I guess in the case of k=1005-2000, that *would* be a much bigger n-range gap than in the case of the 2nd Drive. Okay, let's do it as 2 separate drives then.

As for which thread to unsticky: Hmm, I can think of one right off the bat. How about we unsticky the "Automated Primaility Testing with LLRnet" thread, and instead put a link to it in the "Come Join Us!" thread? After all, its prime usefulness (no pun intended) is for new users, most of whom will be reading the "Come Join Us" thread first anyway.

We should definitely leave News and Teams stickied; those need to be up where people can see them. Ditto for "LLRnet servers for NPLB" and "Report all primes here". Now all that's left are the team drive threads (we'll have 4 of them at first, and a 5th when we add n>350K for k=1005-2000), the individual-k drive, and the doublecheck drive. Possibly the doublecheck drive could be unstickied? Or would that make it too easily fall by the wayside?

Anyway, just a couple ideas.

Max

[Flatlander]

Having lots of stickies is a good thing isn't it? Easy access to everything considered important.

[gd_barnes]

Originally Posted by mdettweiler

Hmm...I see what you mean. Yeah, I guess in the case of k=1005-2000, that *would* be a much bigger n-range gap than in the case of the 2nd Drive. Okay, let's do it as 2 separate drives then.

As for which thread to unsticky: Hmm, I can think of one right off the bat. How about we unsticky the "Automated Primaility Testing with LLRnet" thread, and instead put a link to it in the "Come Join Us!" thread? After all, its prime usefulness (no pun intended) is for new users, most of whom will be reading the "Come Join Us" thread first anyway.

We should definitely leave News and Teams stickied; those need to be up where people can see them. Ditto for "LLRnet servers for NPLB" and "Report all primes here". Now all that's left are the team drive threads (we'll have 4 of them at first, and a 5th when we add n>350K for k=1005-2000), the individual-k drive, and the doublecheck drive. Possibly the doublecheck drive could be unstickied? Or would that make it too easily fall by the wayside?

Anyway, just a couple ideas.

Max

OK, that's a good idea on "Automated Primaility Testing with LLRnet" thread. Can you add a link to it in the "Come Join Us" thread and unsticky it whenever you get a chance?


Thanks,
Gary

[gd_barnes]

Originally Posted by Flatlander

Having lots of stickies is a good thing isn't it? Easy access to everything considered important.

Yes and no. With too many of them, they are off the bottom of many people's screens. I've found that it is easy to miss posts if they aren't on my first screen of threads.

[gd_barnes]

Originally Posted by henryzz

to help solve the stickies problem would it be worth having one stickied thread that has links to several threads that would be sticked otherwise

I'm not so fond of that idea. I think we definitely want the drive threads stickied for visibility. We'd almost have to have a "look here first" thread with the links followed by an explanation as to what all of the important links are since they would frequently drop below the first page of threads on most people's screens. I feel our main drives would lose some visibility that way.

Thanks for the idea though.