Sieved files/sieving coordination

[gd_barnes]

This thread is to coordinate any sieving efforts at Conjectures 'R Us and to post your sieved files that were not searched or were not fully searched for any reason. There will also be team sieving and search efforts and sieved files can be posted here that can be split up and searched by a group of us at the higher search ranges.

There are 2 remaining fully-sieved files as follows:
n=62K-100K for Riesel base 13; 1 k left to prove conjecture.
n=76.5K-100K for Sierp base 17; 3 k's left to prove conjecture.

The 100K limit is high enough for a top-5000 prime and you could score extra points for finding the last prime or next-to-last prime that proves a conjecture!

Edit: Remaining sieved files are now on the Reservations web pages. The attachment to this message can be ignored. I cannot delete an attachment nor edit it in place without deleting the entire post. If you'd like to test one of the above, please get them from the Reservations pages.

Edit by Max: Actually, you can delete attachments, you just have to use "Go Advanced" to see all options. I've removed the attachment, which by now (8/30/09) is even more out of date than it was as of Gary's edit.


Gary

[gd_barnes]

I have just started sieving for team drive #1 on Sierp. base 16. I decided to go ahead and sieve all 57 k's remaining from n=25K-200K. I realize that some are reserved to n=100K so I'm removing the portion of those k's below that limit.

If anyone wants to help with sieving, let me know. As a wild guess for this large of a range and # of k's, we'll need to sieve to ~P=6T.

If the people who have their k's reserved to n=100K want to go higher or if anyone wants to reserve a new k starting from n=25K, then that's fine. Let me know and I'll forward you the portion of the big sieved file that you need.

If anyone has an opinion on k's to leave available outside of the team drive for individual searching, let me know. Right now, I'm thinking we'll probably leave about 8-10 k's available but sieved files will be available on all of them.


Gary

[mdettweiler]

Originally Posted by gd_barnes

I have just started sieving for team drive #1 on Sierp. base 16. I decided to go ahead and sieve all 57 k's remaining from n=25K-200K. I realize that some are reserved to n=100K so I'm removing the portion of those k's below that limit.

If anyone wants to help with sieving, let me know. As a wild guess for this large of a range and # of k's, we'll need to sieve to ~P=6T.

If the people who have their k's reserved to n=100K want to go higher or if anyone wants to reserve a new k starting from n=25K, then that's fine. Let me know and I'll forward you the portion of the big sieved file that you need.

If anyone has an opinion on k's to leave available outside of the team drive for individual searching, let me know. Right now, I'm thinking we'll probably leave about 8-10 k's available but sieved files will be available on all of them.


Gary

Does this include the two k's I'm working on? If so, what happens if I find a prime before I reach n=100k? Does that mean that your sieving work for those k's will have been wasted?

FYI: If I reach n=100k on my two k's without finding a prime, I'll release them, so you can go ahead include them in a drive from n=100k on.

[gd_barnes]

Originally Posted by Anonymous

Does this include the two k's I'm working on? If so, what happens if I find a prime before I reach n=100k? Does that mean that your sieving work for those k's will have been wasted?

FYI: If I reach n=100k on my two k's without finding a prime, I'll release them, so you can go ahead include them in a drive from n=100k on.

Yes it does but only from n=100K-200K. The same thing for the other k that has been reserved. If you or anyone else finds a prime on a k that I'm sieving to a high range, I'll just use srfile to delete the k from the big sieve.

The issue of wasted sieving effort affects all conjecture projects. There's no way around it while still maintaining efficient sieving. If you sieve n=25K-100K and find a prime at n=25001, then the sieving effort is technically wasted if you don't use it. That said, if we're in top-5000 territory, let's say someone is sieving n=100K-200K and they find a prime at n=100001 base 16, then I'd encourage them to continue if what they're after is top-5000 primes, even though it doesn't contribute to this effort. (And if we become an official prime-search project, any top-5000 primes DO add to our score at Prof. Keller's site and I'm all for that!) But if people want to work on proving the conjecture then they can stop and go on to the next k.

This technically makes the mathematical case for stopping sieving before sieving is removing n's as fast as LLR searches them but calculating that would become very complex quickly. I've been sieving until it is removing them as fast as they LLR but I would guess optimum in an effort like this is closer to 80-90% as fast.

Also, you reserved the k's. There's no problem with you taking them above n=100K yourself and just getting that portion of the big sieved file from me or whomever is helping with sieving. By the time you get to n=100K, it's likely we'd be close to having sieved them far enough.

This sieve is intended for both personal use by anyone who wants it and for a team drive, which will benefit everyone as a whole. For that matter, if a majority of people are against a team drive, then I'll just post the big sieved file and people can reserve k's individually and search them using the applicable portion of the file.

Flexibility with everyone's resources and tastes is the key in this and any prime-searching effort.


Gary

[mdettweiler]

Originally Posted by gd_barnes

Yes it does but only from n=100K-200K. The same thing for the other k that has been reserved. If you or anyone else finds a prime on a k that I'm sieving to a high range, I'll just use srfile to delete the k from the big sieve.

The issue of wasted sieving effort affects all conjecture projects. There's no way around it while still maintaining efficient sieving. If you sieve n=25K-100K and find a prime at n=25001, then the sieving effort is technically wasted if you don't use it. That said, if we're in top-5000 territory, let's say someone is sieving n=100K-200K and they find a prime at n=100001 base 16, then I'd encourage them to continue if what they're after is top-5000 primes, even though it doesn't contribute to this effort. (And if we become an official prime-search project, any top-5000 primes DO add to our score at Prof. Keller's site and I'm all for that!) But if people want to work on proving the conjecture then they can stop and go on to the next k.

This technically makes the mathematical case for stopping sieving before sieving is removing n's as fast as LLR searches them but calculating that would become very complex quickly. I've been sieving until it is removing them as fast as they LLR but I would guess optimum in an effort like this is closer to 80-90% as fast.

Also, you reserved the k's. There's no problem with you taking them above n=100K yourself and just getting that portion of the big sieved file from me or whomever is helping with sieving. By the time you get to n=100K, it's likely we'd be close to having sieved them far enough.

This sieve is intended for both personal use by anyone who wants it and for a team drive, which will benefit everyone as a whole. For that matter, if a majority of people are against a team drive, then I'll just post the big sieved file and people can reserve k's individually and search them using the applicable portion of the file.

Flexibility with everyone's resources and tastes is the key in this and any prime-searching effort.


Gary

I get it now. Thanks!

As for my reserved k's: I know I could continue to search them higher, but I'm thinking that once I reach n=100k, I may as well just crunch for the team drive--so, when I reach n=100k (assuming that I don't find any primes, which would of course change the dynamics of the situation entirely), I'll release the two k's and instead work on the team drive. My crunching power is limited, so in some ways team drives are better for me than reserving individual k's.

[gd_barnes]

Sieving will reach P=200G on all 56 remaining k's for Sierp base 16 to n=200K on two cores by early Monday morning (CST US). I just now deleted the sequence for k=7773, which tcadigan found a prime on.

Feel free to still reserve any k for individual work and I'll be glad to provide you with that portion of my big sieve. Tcadigan, if you haven't already sieved that far on your newly reserved k=6663 and want a file sieved to P=200G for n=25K-200K, let me know.


Gary

[gd_barnes]

Sieving will reach P=400G on all 55 remaining k's for Sierp base 16 to n=200K on two cores by late Wed. night (CST US).

Sometime on Thursday, I will separate the reservations page into two pages, one each for Riesel and Sierpinksi, and post a separate sieved file for each Sierp Base 16 k as well as available sieved files for other bases.

People are welcome to reserve any k for individual testing. I will be posting the entire range of n=25K-200K in the sieved files but I would only suggest testing them to n=100K for now. If that is reached without finding a prime, it would be more efficient to wait until sieving has progressed further before continuing.

I've lowered what I think to be the optimum sieve limit now to P=2T. Once that is reached, we'll start a team drive of most of the remaining unreserved k's; leaving a few for individual efforts.


Gary

[gd_barnes]

Tcadigan has kindly posted his fully-sieved file for 64494*4^n+1 for n=885.4K-1M. It is attached below. Let me know if you are interested in searching it and I'll show it as reserved.

Note that this is part of the Sierp Base 4 project in this thread that we are helping coordinate. If you find a prime on it, please report it under that project when creating a prover code on the top-5000 site.


Gary

Edit by Max (8/30/09): cleaned up now-unnecessary attachment

[mdettweiler]

Originally Posted by gd_barnes

Tcadigan has kindly posted his fully-sieved file for 64494*4^n+1 for n=885.4K-1M. It is attached below. Let me know if you are interested in searching it and I'll show it as reserved.

Note that this is part of the Sierp Base 4 project in this thread that we are helping coordinate. If you find a prime on it, please report it under that project when creating a prover code on the top-5000 site.


Gary

Because of the size of the numbers in that sieve file, it would be a pretty hefty effort for any individual cruncher; however, it would be perfect for an LLRNet server, which would allow "small fry" to participate in it (since LLRNet deals in chunks of 1 candidate, rather than large chunks as manual LLRing is usually done in)--although of course "heavy hitters" are just as well suited to LLRNet, too. I could help run an LLRNet server, as long as someone else provides the server (with telnet/SSH access, of course). Any computer that's on 24/7 and has broadband access will do. The load should be small enough that the server could simultaneously be used for crunching, so, for example, a dedicated crunching machine would do as long as it has broadband Internet connectivity. Even though it would probably have a dynamic IP if it's on a residential connection, that can be fixed easily with dynamic DNS.

Let me know what you think!

[gd_barnes]

Originally Posted by Anonymous

Because of the size of the numbers in that sieve file, it would be a pretty hefty effort for any individual cruncher; however, it would be perfect for an LLRNet server, which would allow "small fry" to participate in it (since LLRNet deals in chunks of 1 candidate, rather than large chunks as manual LLRing is usually done in)--although of course "heavy hitters" are just as well suited to LLRNet, too. I could help run an LLRNet server, as long as someone else provides the server (with telnet/SSH access, of course). Any computer that's on 24/7 and has broadband access will do. The load should be small enough that the server could simultaneously be used for crunching, so, for example, a dedicated crunching machine would do as long as it has broadband Internet connectivity. Even though it would probably have a dynamic IP if it's on a residential connection, that can be fixed easily with dynamic DNS.

Let me know what you think!

Agreed on the effort! I'm in favor of this; whatever gets the job done and doesn't step on toes!

But before embarking on it, I'd like to give it about a week for others to pick it up individually first. Additionally, since this is Sierp Base 4, which has it's own project that we are helping coordinate, I'd like to get Jean Penne's OK on it. He is searching all of the other k's for that project currently as well as the Riesel Base 4 k's for this effort.

Jean, do you have the resources to add the above attached sieved file to your testing for Sierp base 4? If not, are you OK with us using an LLRNet server for testing?

One question, Anon, for both this and for our project's future use: If we use an LLRNet server, will the Sierp Base 4 project still get full credit? I'm thinking that the originator of that project would not be in favor of using an LLRNet server if the credit for a huge prime find had to be shared.

Since 'Conjectures 'R Us' is not yet a recognized project on the top-5000 site, it wouldn't matter if it was our effort at this point but it could matter in the future.



Thanks,
Gary

[mdettweiler]

Gary, I've split your post into mulitple parts so I can respond to them individually.

Originally Posted by gd_barnes

Agreed on the effort! I'm in favor of this; whatever gets the job done and doesn't step on toes!

But before embarking on it, I'd like to give it about a week for others to pick it up individually first. Additionally, since this is Sierp Base 4, which has it's own project that we are helping coordinate, I'd like to get Jean Penne's OK on it. He is searching all of the other k's for that project currently as well as the Riesel Base 4 k's for this effort.

Okay, that's fine with me. In the meantime, we might want to see about finding an always-on computer that can run the LLRNet server--Gary, would any of your crunchers possibly work? Even if running this Sierpinski Base 4 k on it fell through, there's plenty of other bases/k's that would be great for LLRNet. (For bases that are powers of 2, you simply use the stock LLRNet client; for other bases, you can use the version modified by the Sierpinski/Riesel Base 5 project.)

Jean, do you have the resources to add the above attached sieved file to your testing for Sierp base 4? If not, are you OK with us using an LLRNet server for testing?

One question, Anon, for both this and for our project's future use: If we use an LLRNet server, will the Sierp Base 4 project still get full credit? I'm thinking that the originator of that project would not be in favor of using an LLRNet server if the credit for a huge prime find had to be shared.

Yes, LLRNet doesn't affect the reporting of a prime--it's reported the same as if it was done with manual LLR. Thus, the reporting would be exactly the same as if the prime was found by someone who was testing those numbers normally.

Since 'Conjectures 'R Us' is not yet a recognized project on the top-5000 site, it wouldn't matter if it was our effort at this point but it could matter in the future.



Thanks,
Gary