BOINC?

[ValerieVonck]

Originally Posted by pacionet

I didn't know LLRnet , now I found some informations. Yes, I think it would be somthing similar but to reduce the traffic over the net it would be better to parse lresult.txt before sending it (sending only the primes found).

Where can we find LLRnet client and server source code ? (Maybe we can modify it )

You don't have to parse lresults.txt. LLRNet Server side will check this.
Goto RieselSieve => download for the client /server(I think).
If you don't find & for configuration issues please contact Lee Stephens (B2)

Regards
C.

[pacionet]

Originally Posted by CedricVonck

You don't have to parse lresults.txt. LLRNet Server side will check this.
Goto RieselSieve => download for the client /server(I think).
If you don't find & for configuration issues please contact Lee Stephens (B2)

Regards
C.

Is there anybody which can look about this ? I am busy in these days, I have no much time to spend ...

[KEP]

Fellow crunchers, things have begun progressing. This day, I recieved a message from Rytis, saying that he has decided to implement the LLR twinprimesearch, and as I checked today the TPS webpage, I could see PrimeGrid has reserved 1 5 m range, and on the primegrid page, the LLR application has been prepared. I know that Rytis will carry out some test the following week, and then he might begin to really load a lot of work for LLR'ing in favor of TPS project of course. He has promised me some feedback, but if this really turns out succesfull (which I've no doubt it will), out project will soon be in need of sievers to ensure a steady workflow, in stead of crunchers, since there is approximately 3,000 active hosts supporting primegrid :)

By the way, can anyone tell me how the propability of finding a twinprime is calculated?

Regards

KEP

[pacionet]

OK, but they will reserve ranges manually from our website ?

About the pre-sieved ranges I think there is no problem because gribozavr is sieving till 25G and I think that will be no problem to release a large range for primegrid.

[KEP]

Originally Posted by pacionet

OK, but they will reserve ranges manually from our website ?

About the pre-sieved ranges I think there is no problem because gribozavr is sieving till 25G and I think that will be no problem to release a large range for primegrid.

I had a conversation with PrimeGrid about this, and I can only see this as an option. If you can release at least 10 G then it would be enough to keep the people at PrimeGrid busy for quite some days, since it would yield approximately 2,000,000 candidates, and thereby give 4,000,000 results to work with for the primers, which is about 1500 results per active computer, which again means that there is work for about 2 whole days. I believe that Rytis is reserving manually on the TPS webpage, the range that he feels like wanting to offer help with solving. At the moment he wants to run the project as a subproject to the primary task at PrimeGrid, which is to find as many small prime numbers as possible, to faster break the RSA challenges later on, so lets all see later on, how much work he will contribute :)

Regards!

[MooMoo2]

Originally Posted by KEP

there is approximately 3,000 active hosts supporting primegrid :)

By the way, can anyone tell me how the propability of finding a twinprime is calculated?

Regards

KEP

The chance of a random integer being prime is 1 in log2(x). We are currently looking for twins that are about 58,700 digits long, and the chance of a random 58,700 digit number being prime is about 1 in 136,000.

However, k*2^195000-1 is always an odd number, so with no sieving, the chance of a random k being prime is 1 in 68,000 (136000 divided by 2). 68,000^2 is 4,624,000,000, so the chance of a k being a twin is 1 in 4,624,000,000.

Fortunately, sieving eliminates many of these candidates. Each 5M file has 2.5M candidates, but sieving only leaves slightly more than 1,800 candidates per file. 2.5M/1.8K is about 1400, which means sieving reduces the amount of candidates by a factor of 1400.

4.6G/1400 is about 3.3M. Therefore, the chance that a random k from the sieving file gives a twin prime is 1 in 3.3M.

In my first post in this thread, I said that the odds were around 1 in 8M. This was because I used the old sieve files as estimates. At that time, each 5M file had over 2500 candidates, as sieving was still beginning.

----------------

I'll be really happy if primegrid helps us out. Assuming half of the computers in primegrid join us, we should find a twin within a week

[Rytis]

OK, so KEP guided me to this thread, and here I am :)

My initial statement about 300€ initial deployment costs were stated in case I would implement the system on your servers; however, I decided to help you a little and deploy the application on my server. (Just to let you know, I really like gifts, and running a pretty high-end server in a datacenter costs money, so I (=PrimeGrid) would really like some money, if that could be found and spared. Not that I'm stating this is necessary)

So far, I deployed a bit modified version of LLR to PrimeGrid (but as of now there is no work, so don't bother trying to get some). The modifications are only to report progress to a text file (so that BOINC can display it) and to quit after work done and not wait for more work. The modifications were done with help of RieselSieve guy Bryan. I helped him to deploy BOINCsieve on RieselSieve, so it was his turn to help me ;)

However, legacy application wrapper needs more work, because LLR can't find the input file, as of current version. Hopefully I am able to fix the problem today in the evening (Europe time).

Now I have reserved a 5M range for tests (925M-930M), sieved it on my PC to about 3000 candidates (for testing, there is no need to sieve more, I want to have a bit more test workunits without reserving a huge range). Workunits as of now will be only one LLR test (however, you can get more than one workunit at a time). Later I will implement test batching (it needs more work on the wrapper) so workunit duration will be longer.

Of course, I need to write work generator and assimilator (returned work processing) first. Now I insert work into the database manually, and there was no successful work returned yet, so... :)

If you join PrimeGrid to see LLR tests, go to project specific preferences in your account and choose to run beta apps - LLR is one of them. If there is no LLR work, you will get standard PrimeGrid app to work on.

And, feel free to ask questions :)

[pacionet]

As soon as possible (when gribozavr send me (thanks gribozavr for your work !!!)) I'll upload on the pre-sieved page of the site all ranges till 11G in the form (1000-2000M, 2000-3000M , etc) and , I hope, in the form (1005-1010, 1010-1015 and so on) The first for people who wants to reserve huge range and the second for normal single user.

Maybe, when all the users of primegrid join the project entirely, Rytis can reserve for them a huge range (some G's) .

After some time, I think that we can all become primegrid users for LLR BOINC test and I can transform the twinprimesearch project in a sieving project for primegrid.

Do you agree ?

Andrea

[gribozavr]

You see, sieving is still going on at a rate of about 15-30 sec/k. I don't want to release big ranges if they would just sit on server for a long time. But if PrimeGrid will LLR all that in comparable time (so there would be very little benefit from sieving) -- it's great.

[pacionet]

Originally Posted by gribozavr

You see, sieving is still going on at a rate of about 15-30 sec/k. I don't want to release big ranges if they would just sit on server for a long time. But if PrimeGrid will LLR all that in comparable time (so there would be very little benefit from sieving) -- it's great.

Rytis, tell us how many Giga (or Millions) you think to reserve and we release them.

[Rytis]

First let me create the application and test how fast can we go :)