BOINC?

[MooMoo2]

Originally Posted by pacionet

Wow !
Gribozavr we need next pre-sieved ranges 1100-1200M !!!!

Thanks
Andrea

You (or Victor) can send me the "All in one" file with all the 100M range (for Primegrid) ?
I 'll post them here, in "All in one", bottom of the page:

http://www.twinprimesearch.org/pre-sieved_ranges.htm

Thanks
Andrea

At this rate, we'll find a record-size twin by the end of this year

[biwema]

Originally Posted by KEP

I just started a for fun test, to see exactly how long it would take, and apparently it will take 99 milliseconds to do 1 iteration, which means that it on my machine will take around 38 days ± ½ day, so as it turns out, LLR is faster on 10 M digit candidates than MPrime is, even though the exact is opposite with the n we are currently testing :) Yeehhh my bad :) Apparently the range though is pretty fast to sieve, so plenty of work should be around for long time :) I might give it a try see how it goes with the 25 G range, if I try to sieve it to 100 Billion :)

finding a 10 Megadigit twin is really a hard task.

Expect searching a range of 400000G - 500000G and testing almost 100 Billion Candidates (during that time you might find 300000 non-twin 10M digit primes), depending on your sieving depth.

Sieving a range of 500G needs about 1.5 G Ram, so better wait until NewPGen Supports a huge Memory Size, before start sieving. Then sieving much more than 1000 CPU years might be appropriate.

Nevertheless, Good luck...

[KEP]

Originally Posted by MooooMoo

At this rate, we'll find a record-size twin by the end of this year

Lets hope so, I think it could be cool to hit the first twin prime, before the end of the year, but still even though Rytis has automatized the LLR job creation process, it will require a load of users from the primegrid network, to start requesting LLR wu, and no primegen WU for those of us doing a hard effort to find the next twin :) No offence Rytis...

Originally Posted by biwema

finding a 10 Megadigit twin is really a hard task.

Expect searching a range of 400000G - 500000G and testing almost 100 Billion Candidates (during that time you might find 300000 non-twin 10M digit primes), depending on your sieving depth.

Sieving a range of 500G needs about 1.5 G Ram, so better wait until NewPGen Supports a huge Memory Size, before start sieving. Then sieving much more than 1000 CPU years might be appropriate.

Nevertheless, Good luck...

I couldn't sieve it, actually it has been sieved to 1 billion, but the program made some weird mistakes, like sieving out of the range requested, and when it had to save the work, it crashed :) So I'm not gonna sieve it anyway. But maybe you can tell me how many G is needed to be sieved, before one can say for sure to be able to find a 1 million digit twinprime?

[gribozavr]

Rytis, I've been watching http://www.primegrid.com/orig/range_grid.php and http://www.primegrid.com/?charts (Unsent LLR results) and can't yet understand some terms.

What are "unprepared tasks"?
"Done tasks" means that the number has been tested and the server got the result?
What are "Unsent LLR results"?

[Rytis]

When I initially insert new ranges into the database, they are not created as BOINC workunits, but reside in the temporary llr table. Range grid shows them as red rows, and the actual number of tasks not converted into BOINC workunits is shown in unprepared tasks column.

Later when BOINC client requests work for LLR, generator is launched, creating two results (in BOINC terms, a result is a single instance of workunit, does not matter if returned by the user or not yet sent. I know, confusing) for every row in LLR table (technically not exactly like described, but let's simplify things not going too far from the idea). The generator tries to keep 600 unsent results (for now) to be able to send work any time a user requests for it. This is what you see in the charts table as unsernt LLR results.

Back to range grid. A workunit is considered complete only when two (or more, in some cases) matching results have been returned. Then it is checked for primes and marked as complete (also, BOINC credit is granted). Such complete results are shown in done tasks column.

I hope it got clearer, let me know if you need additional explanations.

[biwema]

Originally Posted by KEP

But maybe you can tell me how many G is needed to be sieved, before one can say for sure to be able to find a 1 million digit twinprime?

- You are for sure able to find a Megadigit prime by testing one candidate, if you are lucky.

- You will never be sure that you really will find one. (independant from the size of the range)

I assume you want to know when it is likely to find one:
After sieving 7000 G - 7500G you will have the same chance to find a megadigit twin (n=3330000) as twinprimesearch has it now with n=195000 an a range of 25G.

I will post more accurate information later.

[KEP]

Originally Posted by biwema

- You are for sure able to find a Megadigit prime by testing one candidate, if you are lucky.

- You will never be sure that you really will find one. (independant from the size of the range)

I assume you want to know when it is likely to find one:
After sieving 7000 G - 7500G you will have the same chance to find a megadigit twin (n=3330000) as twinprimesearch has it now with n=195000 an a range of 25G.

I will post more accurate information later.

Thanks, I'll awaite that accurate information with patience and exitement To those who doesn't check back to this thread, before or mayby never again does it, I'll say, Merry Christmas and may you all have a Happy New Year

[smh]

Originally Posted by Rytis

A workunit is considered complete only when two (or more, in some cases) matching results have been returned.

I think it's a waste of time to check all numbers twice. We're not trying to find ALL primes in this range, just looking for A twin prime.

One more question, How can i make sure i do not get any primegen WU's? I have a spare pc which can do some useful work on the twin prime search, but primegen isn't even worth the electricity to run my pc.

[Rytis]

I do doublechecking because we had people cheating in BOINC. Also, I think that we don't want to miss a twin prime, because that would be waste of computing power - we would need to go much higher to find another one.

As for getting only LLRs, if you have set to run test applications and there are red rows in range_grid page (that is, unsent work), you will get LLRs as they have higher priority.

Originally Posted by Rytis

I do doublechecking because we had people cheating in BOINC. Also, I think that we don't want to miss a twin prime, because that would be waste of computing power - we would need to go much higher to find another one.

But going higher doenst take longer. Tests are the same length. Doublechecking really is a waste of time... it will just take twice the time to find a prime. Because 95% of reasidues or more will match and those are wasted.

[gribozavr]

Maybe doublechecking should be selective based on the user who submitted the result? We trust people on mersenneforum, they won't send fake results -- so their work doesn't need to be doublechecked. You probably trust your top-producers who have many machines running (but after some doublechecking to be sure that the machines are not faulty).