mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Prime Cullen Prime (https://www.mersenneforum.org/forumdisplay.php?f=79)
-   -   Collaboration with PrimeGrid? (https://www.mersenneforum.org/showthread.php?t=8599)

hhh 2007-07-05 14:42

Collaboration with PrimeGrid?
 
[url]www.primegrid.com[/url] is currently starting Woodall and Cullen searches without consideration of the primality of the candidate n's. While Woodall is asready full steam ahead, the Cullen-part isn't online yet because the client needs some adaptations. The start is in about a week or so.

I had some contact with Rytis, the primegrid's admin, and there is the possibility to just release all of our already sieved numbers into their queue before they start with their candidates, and to get residues in no time.

Of course this would mean that we just have to restart sieving, and the perspective for the future would be completely changed.

What do you think?

Yours H.

Citrix 2007-07-05 15:48

[QUOTE=hhh;109655][url]www.primegrid.com[/url] is currently starting Woodall and Cullen searches without consideration of the primality of the candidate n's. While Woodall is asready full steam ahead, the Cullen-part isn't online yet because the client needs some adaptations. The start is in about a week or so.

I had some contact with Rytis, the primegrid's admin, and there is the possibility to just release all of our already sieved numbers into their queue before they start with their candidates, and to get residues in no time.

Of course this would mean that we just have to restart sieving, and the perspective for the future would be completely changed.

What do you think?

Yours H.[/QUOTE]


Well they could remove all prime n's from their queue and we can make sure that we always stay ahead of them. This way no primes would be missed.

Siemelink 2007-07-05 19:22

[QUOTE=hhh;109655][url]www.primegrid.com[/url] is currently starting Woodall and Cullen searches without consideration of the primality of the candidate n's. While Woodall is asready full steam ahead, the Cullen-part isn't online yet because the client needs some adaptations. The start is in about a week or so.

What do you think?

Yours H.[/QUOTE]

I think that a dual approach would leave people of different temperament to choose what they are after. Mass effort vs tailored approach.
As for me, my Woodall effort is drawing to a close. I'll finish soon with the range 1.2M until 1.6M. I have my eyes on a different project so I welcome someone else effort on Woodalls.

Willem.

hhh 2007-07-05 20:04

[QUOTE=Citrix;109669]Well they could remove all prime n's from their queue and we can make sure that we always stay ahead of them. This way no primes would be missed.[/QUOTE]

Can we? PrimeGrid is no baby framework.
On the other hand: if we stay ahead, we can let them do the doublecheck as well. H.

Phased_Array 2007-07-05 21:55

I don't think this project can stay ahead of PrimeGrid. With just a few early testers they were completing tests at a rate of 1000 K's/hour (lower K/N values than on this project though).

Best bet if you want to keep this project going (and I hope you do) is ask them (Rytis and/or John [jmbazek]) to remove the ranges your testing. Hopefully they will comply with the request.

If you do decide to continue with the project, I would like to do a little LLR work if possible. Good Luck!

hhh 2007-07-06 07:05

[QUOTE=Phased_Array;109715]
Best bet if you want to keep this project going (and I hope you do) is ask them (Rytis and/or John [jmbazek]) to remove the ranges your testing. Hopefully they will comply with the request. [/QUOTE]

They proposed that already. But wouldn't it be contra-intuitive to keep doing tests that would have been done faster anyway? The whole idea about PCP was to do the prime tests before.

What about this scenario: We give them some candidates, say up to 4M, and keep a few for us. And we restart sieving.

If they still overhaul us, we reiterate the process. If not, well, then fine. This way, people have the choice to do testing in advance to the masses, and the masses do some work in advance they would have done later anyways. IMO, that would be a good compromise.

I think it would be narrow-minded just to turn a blind eye to something that IS THERE.

Yours H.

rogue 2007-07-06 13:04

Who runs PrimeGrid? It would have been nice if they had contacted either me or Ray Ballinger since the search has been coordinated on [url]www.prothsearch.net[/url] for years. There are people who are currently working on ranges and it isn't particularly nice of them to "poach" from another project.

Phased_Array 2007-07-06 15:41

They "claim" to have asked around ("if anyone was working on these ranges"). I have yet to see any mention of this by them in of the usual places.

hhh 2007-07-06 15:54

Well, they asked me, for instance. They should have asked you, though, rogue. Hmm. H.

geoff 2007-07-07 00:38

Since primegrid is sieving all candidates in a larger range, it might take them some time to reach the depth that we have sieved to, so I would like to keep LLR testing our candidates at least until they catch up, if that doesn't complicate things too much.

After that it would be good if we could combine our resources in sieving, but still have some way of focusing our LLR efforts on the Cullens with prime exponent. I don't know how best to do that, we certainly don't want to reserve the prime exponents for seperate testing if that would delay them being tested along with the others.

rogue 2007-07-09 16:47

I have been working with the folks at PrimeGrid so all issues will be worked out shortly. AFAIK, they have not "poached" work from anybody as they contacted other searchers on their own.

SB2 2007-07-09 17:00

I noticed they have started sending out Woodall candidates again, so I guess everything has worked out for all involved?

hhh 2007-07-09 20:03

[QUOTE=rogue;109936]I have been working with the folks at PrimeGrid so all issues will be worked out shortly. AFAIK, they have not "poached" work from anybody as they contacted other searchers on their own.[/QUOTE]

Good to hear.

[QUOTE=SB2;109936] I noticed they have started sending out Woodall candidates again, so I guess everything has worked out for all involved?[/QUOTE]

As far as I understood, Woodal is working, and Cullen has some minor Client issues yet to solve.

Does anybody take offence if I give them the numbers from 2.5M-3M as a test? So we could see how quick they are, and later how fast they do with Cullen.

I will provide some new sieving soon, as well.

Yours H.

SB2 2007-07-09 22:52

[QUOTE=hhh;109954]Good to hear.



As far as I understood, Woodal is working, and Cullen has some minor Client issues yet to solve.

Does anybody take offence if I give them the numbers from 2.5M-3M as a test? So we could see how quick they are, and later how fast they do with Cullen.

I will provide some new sieving soon, as well.

Yours H.[/QUOTE]

How many candidates are there from the area we are LLR testing to 2.5M? Enough to keep us busy LLR'ing while sieving a higher range?

It is your project, but if we wish to continue the project as envisioned, I fail to see the reasoning behind giving them the smaller/faster candidates when it is obvious they have far more resources available, therefore more capable to do the higher ranges. After all thier plan is to test upto 10M. :confused:

hhh 2007-07-11 09:22

[QUOTE=SB2;109962]How many candidates are there from the area we are LLR testing to 2.5M? Enough to keep us busy LLR'ing while sieving a higher range?

It is your project, but if we wish to continue the project as envisioned, I fail to see the reasoning behind giving them the smaller/faster candidates when it is obvious they have far more resources available, therefore more capable to do the higher ranges. After all thier plan is to test upto 10M. :confused:[/QUOTE]

Actually it is your project, or at least ours.

We had about 3800 candidates in the 1.5M-5M-range, uniformly distributed. the 5M-tests are estimated to take about one day each.

From your response, I guess it is the best to stick to geoffs proposal, which is wait and see how things develop, while of course helping PG with factors and so on.

Meanwhile, I think we are ready for the near future: Sieving is online again, I have fully P-1ed LLR candidates up to 3M (need only to post them), and P-1 work is still available as well.

Cheers, Yours H.

geoff 2007-09-07 03:24

Primegrid is catching up. PCP LLR has almost reached n=3,000,000, but at this rate Primegrid will probably overtake well before n=4,000,000.

I would like to keep LLR testing up to n=3,000,000, but I don't think I'll continue beyond that.

hhh 2007-09-07 23:49

[QUOTE=geoff;113748]Primegrid is catching up. PCP LLR has almost reached n=3,000,000, but at this rate Primegrid will probably overtake well before n=4,000,000.

I would like to keep LLR testing up to n=3,000,000, but I don't think I'll continue beyond that.[/QUOTE]

It is indeed, see this page

[url]http://www.primegrid.com/range_grid.php[/url]

and that's why I suggest leaving PrimeCullenPrime just die in silence and dignity.

In fact there isn't even a point in continuing LLR, as they already run the doublecheck immediately.

If anybody sees a future to this project, he may speak, or remain silent forever. Or speak later, indeed. H.

Citrix 2007-09-08 00:10

How far is primegrid going to PRP? Are they only going to go to 3M? How far have they sieved their ranges? What program do they use to sieve?

jasong 2007-09-08 00:43

On that note, I'll unreserve the ranges I promised I'd continue doing.(I was a lying bastard even before I saw this thread :( )

It would be nice if PrimeGrid would give people the power to specifically test prime Cullen ns, or whatever they're called.

/me goes to the proper thread to unreserve his ranges.

BlisteringSheep 2007-09-08 02:41

[QUOTE=jasong;113825]It would be nice if PrimeGrid would give people the power to specifically test prime Cullen ns, or whatever they're called.[/QUOTE]
If I understand what you're asking for, it's already provided (and has been since they added Woodall & Cullen). On the [URL="http://www.primegrid.com/prefs.php?subset=project"]PrimeGrid preferences page[/URL], you can chose to which type(s) of work you want, from Primegen, Twin Prime Search, Woodall Prime Search and Cullen Prime Search. You can do this globally or by BOINC venue (global, home, school, work).

And prior to that, you could pick TPS instead of Primegen by enabling beta applications.

:alex:

Citrix 2007-09-08 20:17

[QUOTE=jasong;113825]

It would be nice if PrimeGrid would give people the power to specifically test prime Cullen ns, or whatever they're called.

[/QUOTE]

Perhaps we can set our own BOINC projects (So all non-GIMPS but mersenne forum projects) can get some BOINC users and speed up things. This way it might be possible for us to stay ahead of them. Does anyone know how to set up a BOINC project? :smile:

hhh 2007-09-10 14:29

As far as I'm concerned, I neither have the power, the ressources nor the knowledge for doing so.

And I think, as a general guideline, it is better to leave to the king what belongs to the king: Primegrid's runner Rytis was the one who made PerlBoinc, as far as I know. For him, it would be like a fingersnip to implement a special choice for prime n. Well, I guess it would. So, I think our time is better spent convincing him to do so. On the other hand, there are for sure arguments against doing so, too, that I don't know.

I think pretty much that PrimeCullenPrime is done, and I'm glad that somebody else is searching for the same thing and more faster than we did.

Yours H.

jasong 2007-09-10 22:15

Unfortunately, for me, only Windows is a choice in that project, so I only have one computer I can use, and it's committed to another project for a good long while.

geoff 2007-09-11 04:17

If we wanted to ensure a certain milestone was achieved early, say LLR testing to n=10M, it would be logical to start at n=10M and LLR the candidates in reverse order, and meet Primegrid on the way up.

But I am not volunteering to administer such a search :-) I would rather let Primegrid do the hard LLR work, and cheer when they find the first Prime Cullen Prime.

hhh 2007-09-11 06:52

[QUOTE=geoff;114033]I would rather let Primegrid do the hard LLR work, and cheer when they find the first Prime Cullen Prime.[/QUOTE]

That's what I think. If somebody else is doing your work, and better with less effort, you'd be better devoting yourself to something else.

The point of PrimeCullenPrime was to do some search at all, and that worked fine. Almost nobody seemed interested, but we would have arrived to something over the years. PG is doing all that and more in a few months.

For those who want niche projects, there is always Sierpinski base 4 and stuff.

The only useful thing I can imagine we could do is sieving some only prime candidates for a future Prime Grid "prime n" assault. Yet, there are probably better ways for spending our few ressources.

H.


All times are UTC. The time now is 15:01.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.