mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Twin Prime Search (https://www.mersenneforum.org/forumdisplay.php?f=65)
-   -   Next Exponent(s) (https://www.mersenneforum.org/showthread.php?t=6965)

pacionet 2007-01-15 10:04

Next Exponent(s)
 
I offer myself to sieve another exponent (maybe the one following n=333,333 ?) and release periodically the pre-sieved files (like Gribozavr did for n=195,000).

I'll use [url]www.twinprimesearch.org[/url] to upload the pre-sieved files.

I'd like to work alone on this exponent to avoid to transfer big files. Is it possible ?

Moreover everybody should go to PrimeGrid to do LLR testing.

KEP 2007-02-18 12:41

Moomooo and pacionet, just for curiosity, what is the sieving depth, and how many of the 208 G candidates, is left? At the moment, I've just calculated, that there is 348 years of P4 2.553 GHz of computation left on n=333,333, but of course it will be lover since sieving is still going strong, and since we shouldn't expect to go all the 100 G through before a twin is found... again it is only out of curiosity, and I'm not going to offer any sieving help, since I'm busy trying to find primes with LLR at primegrid, since my computer is now again healthy though it has had its ram reduced to the half :geek:

Regards!

KEP

Ps. I lost track of the thread currently about sieving n=500,000 that's why I asked here in this thread :)

pacionet 2007-02-18 14:26

[QUOTE=KEP;98872]Moomooo and pacionet, just for curiosity, what is the sieving depth, and how many of the 208 G candidates, is left? [/QUOTE]

Check in "Sieving discussion forum"

thommy 2008-04-26 17:42

So what's next? Is it Primegrid's 666,666 exponent that is sieved for SG and TP chance or are you palnning a twin only search again, like 500,000 as the exponent. Wouldn't be that efficient. Is there sieving done yet?

MooooMoo 2008-04-26 19:19

[QUOTE=thommy;132221]So what's next? Is it Primegrid's 666,666 exponent that is sieved for SG and TP chance or are you palnning a twin only search again, like 500,000 as the exponent. Wouldn't be that efficient. Is there sieving done yet?[/QUOTE]
Well, we actually have two possible directions for the next exponent:

1.) Test n=500,000 as planned earlier. I think cipher's sieving 1-50G, you'll have to ask him how the progress is on that range.
2.) Change course and follow Gary, Robert, et al's advice on this thread:

[url]http://www.mersenneforum.org/showthread.php?t=8479&page=2[/url]

They're suggesting to sieve a broad k and n range. Sieving will be less efficient, but the LLR tests should not take as long. In that case, the n-range will be from n=450,001 to n=524,288, and the k range will be from k=3 to k=2.8M.

I haven't made any final decisions as to which choice to take. The default is option 1, but I'm open to suggestions.

robert44444uk 2008-04-27 10:42

The purpose of option 2 is to virtually guarantee a twin, whereas with picking a single n this is not possible. I would be interested to see how many computer years such an exercise would take, because, testing the option 2 approach at n=71000 will take quite a bit of time so at n=500000 the time must be astronomical. The current record was achieved through good fortune if we look at the statistics.

gd_barnes 2008-05-03 03:52

Another option might be to 'scale down' a little. I would suggest a "variable-n" approach using the range of n=250K-350K, of course skipping n=333333. There would be little or no top-5000 primes but I think the point here is to find a record twin, not find multitudes of single primes that will quickly drop off the top-5000 list. But only you guys know if finding top-5000 primes adds significant resources to the project.

By my estimate, PrimeGrid will not find an n=666666 twin or SG for 5+ years, even with a huge amount of resources.

I think you want to virtually guarantee that you'll find the largest twin when you do find one. If you choose n=500K or n=450K-524288, it's very possible that might take longer than n=666666 for PrimeGrid depending on how many more resources they have than you.

Just my two cents...

robert44444uk 2008-05-03 12:27

Amphoria got a positive result for the final test using the many n, low k method:

1294767*2^67708+/-1

So this completes all testing of the method... I am now confident it will produce the goods at any level of n.

Given the resources available, which are limited, then I suggest an approach similar to Gary, with a goal of finding the largest twin to date in, say, 3 months.

It is no use setting a five year goal...people will just not stick with it


All times are UTC. The time now is 16:29.

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