mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Conjectures 'R Us (https://www.mersenneforum.org/forumdisplay.php?f=81)
-   -   Recommended bases and efforts (https://www.mersenneforum.org/showthread.php?t=13196)

mdettweiler 2010-05-21 15:34

[quote=gd_barnes;215592]Added 49*476^n-1 for n=25K-100K. The proof of R476 would give us 5 consecutive proven bases in numeric succession for the first time. R472 thru R475 are already proven.[/quote]
Hey Gary, could you possibly sieve that one for me like you did for R181, R182, S140, and S187? The high priority on this one would seem to make it a good next choice for my quad.

While I'm at it, I may as well throw R191, R213, R218, and R236 on the list (again for 25K-100K except R191 which would be 50K-100K). My quad is pretty close to finishing R187 and I'll be ready for another batch of bases soon. This isn't an enormously high priority at the moment since I was planning to do a 1-week or so stint on one of the team drives as a "break" from this little effort once I'm done with the current batch, so if you've got something else loaded up ahead of this right now, feel free to finish that before you start on this.

gd_barnes 2010-05-21 21:25

[quote=mdettweiler;215641]Hey Gary, could you possibly sieve that one for me like you did for R181, R182, S140, and S187? The high priority on this one would seem to make it a good next choice for my quad.

While I'm at it, I may as well throw R191, R213, R218, and R236 on the list (again for 25K-100K except R191 which would be 50K-100K). My quad is pretty close to finishing R187 and I'll be ready for another batch of bases soon. This isn't an enormously high priority at the moment since I was planning to do a 1-week or so stint on one of the team drives as a "break" from this little effort once I'm done with the current batch, so if you've got something else loaded up ahead of this right now, feel free to finish that before you start on this.[/quote]

OK, I can do that. One thing though: The 1k remaining bases 200-250 at n<100K (most/all at n=25K) would be higher priority than the 2k remaining bases 200-250 at n<100K. There's a lot of 1k bases on both sides and they would give you a better chance at a proof. Let me know if you'd rather switch out R213/R218/R236 for some of those. (I assume you'd still keep R476/R191.) In the mean time, I'll start on R476.

I suppose I could put the 1k remaining bases 200-250 at n<100K in the recommended thread here. If I do that, I'll need to look at all bases 200-250 (not just the 1k/2k ones) because I didn't go past base 200 when I originally made this thread other than on those 1k k=2 bases. I'll look at doing that a little later.

mdettweiler 2010-05-21 23:09

[quote=gd_barnes;215665]OK, I can do that. One thing though: The 1k remaining bases 200-250 at n<100K (most/all at n=25K) would be higher priority than the 2k remaining bases 200-250 at n<100K. There's a lot of 1k bases on both sides and they would give you a better chance at a proof. Let me know if you'd rather switch out R213/R218/R236 for some of those. (I assume you'd still keep R476/R191.) In the mean time, I'll start on R476.

I suppose I could put the 1k remaining bases 200-250 at n<100K in the recommended thread here. If I do that, I'll need to look at all bases 200-250 (not just the 1k/2k ones) because I didn't go past base 200 when I originally made this thread other than on those 1k k=2 bases. I'll look at doing that a little later.[/quote]
Okay, yeah, that sounds good. In that case, then, I'll do the following:

R476 from 25K
R191 from 50K
S208 from 50K
S230 from 25K
R214 from 50K
R221 from 50K

That should be enough to keep me busy for a little while. :smile:

gd_barnes 2010-05-22 23:10

Max,

Doin' a little resource planning here...

You caught me at exactly the right moment. I'm just now finished sieving base 22 to P=13T on 3 cores of my I7 and was going to expand that out to all 8 cores today to bring it up to around P=35T-40T in 10-12 days or so, which should be close to optimum for the n=300K-500K range. Once I start that, I don't really want to stop it so that I can easily split the ranges equally across all cores and have them complete at about the same time, so instead I'm going to use all cores to sieve all of your bases at once very quickly before continuing with base 22.

In doing this, would you possibly like to add R248 for n=25K-100K? That would put all current 1k bases <= 250 at n=100K. If not, I'll go ahead and take it. That will give me 7 bases to sieve, which will almost perfectly utilize the I7 for a couple of days.

Doing that I believe would allow me to finish yours and the base 22 sieving just in time for the NPLB rally in just under 2 weeks. Also, I'm very confident those 7 bases would be enough to keep your quad busy until after the rally. If not, I know you were talking about possibly doing a base 16 reservation in there.

One more advantage: If I send you all of the files at once, you should be able to test them all at once, especially if you use a PRPnet server, which I believe is quite adept at handling multiple bases yet knowing when to stop searching a k when a specific k/base combo finds a prime. Doing that saves you quite a bit of time. The only tricky part is sorting the results back out by base but I'm sure that'd be a snap for your automated scripts.


Gary

mdettweiler 2010-05-23 00:44

[quote=gd_barnes;215760]Max,

Doin' a little resource planning here...

You caught me at exactly the right moment. I'm just now finished sieving base 22 to P=13T on 3 cores of my I7 and was going to expand that out to all 8 cores today to bring it up to around P=35T-40T in 10-12 days or so, which should be close to optimum for the n=300K-500K range. Once I start that, I don't really want to stop it so that I can easily split the ranges equally across all cores and have them complete at about the same time, so instead I'm going to use all cores to sieve all of your bases at once very quickly before continuing with base 22.

In doing this, would you possibly like to add R248 for n=25K-100K? That would put all current 1k bases <= 250 at n=100K. If not, I'll go ahead and take it. That will give me 7 bases to sieve, which will almost perfectly utilize the I7 for a couple of days.[/quote]
Sure, why not...another 1k base on the pile shouldnt make much of a difference. :smile:

[quote]Doing that I believe would allow me to finish yours and the base 22 sieving just in time for the NPLB rally in just under 2 weeks. Also, I'm very confident those 7 bases would be enough to keep your quad busy until after the rally. If not, I know you were talking about possibly doing a base 16 reservation in there.[/quote]
Yeah, no big deal either way. As you said, I'll be doing stuff from the team drives in between, so finding filler work shouldn't be a problem.

[quote]One more advantage: If I send you all of the files at once, you should be able to test them all at once, especially if you use a PRPnet server, which I believe is quite adept at handling multiple bases yet knowing when to stop searching a k when a specific k/base combo finds a prime. Doing that saves you quite a bit of time. The only tricky part is sorting the results back out by base but I'm sure that'd be a snap for your automated scripts.[/quote]
Actually, what I was doing is using PRPnet's sortoption=A on the server, which hands out candidates by order of when they were loaded into the server*. I then load in each of the files as they come via prpadmin, which means that the server does them in that order. If I have multiple files to load in at once, they still need to be applied individually anyway, so thus they're done in the order that I load them in.

Nonetheless, correct, it's not a problem for my scripts to sort that out...in fact some of that usually happens at the beginning and end of each range, since when one range finishes it just starts right in on the next and thus they're mixed within the same daily results files. (The only time this [i]doesn't[/i] happen is when the server dries out at the end of a range and stays that way at least until the next daily copy-off.)

[SIZE=1]*Well, essentially that works out to sorting by decimal length within each particular sieve file. That's because files that are individually this small wind up being uploaded to the server so fast that the entire file has the same timestamp (counted in seconds) on it, so the decimal length of the candidates is used as a tiebreaker where applicable. At any rate, that pretty much sorts by n so since the files are sorted by n, the net effect is the same.[/SIZE]

gd_barnes 2010-05-23 05:22

I've added five sequences that have only k=2 remaining to the 1k recommended list. k=2 should always be the fastest searching k so it is favorable to search across all bases.

gd_barnes 2010-06-08 02:24

Added another recommended 1k base that has only k=2 remaining:

2*836^n+1

gd_barnes 2010-06-24 08:38

vmod,

Since you've now completed the recommended range for S55, I have removed it from the recommended bases here and moved all related S55 posts to the bases 33-100 thread. That is what I do when people complete the recommended parts of the bases here.

Nice work!


Gary

gd_barnes 2010-06-28 23:30

I have now added the remaining bases <= 250 that currently have only 2 or 3 k's remaining at n<100K to the 1st post here. We'd like to push these to n=100K to see if we can bring them down to 1 k remaining. For an easy summary of just those bases, see the link towards the bottom of the 1st post. There are plenty of unreserved bases that are at n<=50K.

This is in preparation for a large team effort where we will sieve and load remaining 1k unreserved bases <= 250 that are at n=100K into a PRPnet server.

mdettweiler 2010-06-29 01:31

Gary,

Now that I'm done with my latest crop of 1-k bases (S208 and S230 are done "unofficially" and I see you've marked them as done in the various listings; they'll be processed tomorrow and the results posted then), I'll be ready for a new batch. As you said in the "Conjectures with one k remaining" thread, all 1-k bases <=250 are now at least at n=100K, so the current focus will be on 2-k and 3-k bases; the ones I'm thinking of for my next batch are:

S157, n=25K-100K
S217, n=25K-100K
R103, n=40K-100K
R213, n=25K-100K
R218, n=50K-100K
R236, n=25K-100K

That is, all the remaining bases on the 2- and 3-k bases <100K list. Note that they don't have to be ready right away; right now I've loaded up my PRPnet server with n=20K worth of S6 work to hopefully hold me over until I get back from my upcoming vacation July 4-10. (I'd rather not do bases with only a few k's remaining while I'm away, in case I get really lucky and run out of work prematurely.) So, if you could get the above bases sieved by about July 10 (even a little later should be OK), that would be great.

Thanks,
Max :smile:

gd_barnes 2010-06-29 01:59

[quote=mdettweiler;220120]Gary,

Now that I'm done with my latest crop of 1-k bases (S208 and S230 are done "unofficially" and I see you've marked them as done in the various listings; they'll be processed tomorrow and the results posted then), I'll be ready for a new batch. As you said in the "Conjectures with one k remaining" thread, all 1-k bases <=250 are now at least at n=100K, so the current focus will be on 2-k and 3-k bases; the ones I'm thinking of for my next batch are:

S157, n=25K-100K
S217, n=25K-100K
R103, n=40K-100K
R213, n=25K-100K
R218, n=50K-100K
R236, n=25K-100K

That is, all the remaining bases on the 2- and 3-k bases <100K list. Note that they don't have to be ready right away; right now I've loaded up my PRPnet server with n=20K worth of S6 work to hopefully hold me over until I get back from my upcoming vacation July 4-10. (I'd rather not do bases with only a few k's remaining while I'm away, in case I get really lucky and run out of work prematurely.) So, if you could get the above bases sieved by about July 10 (even a little later should be OK), that would be great.

Thanks,
Max :smile:[/quote]


Very good. I'll show them all as reserved by you. I was hoping that would encourage you to do some of those. It'll be a little while as my I7 is loaded up right now but 12 days should be plenty of time.

Have a good trip. :smile:


All times are UTC. The time now is 23:01.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.