Thanks PG for Sieving, now time to pick a new n?

[Prime95]

Sorry for not bringing this up earlier: PRP iteration timings are about 2.5% faster if you select an exponent that is a multiple of the FFT length.

[Puzzle-Peter]

Originally Posted by Prime95

Sorry for not bringing this up earlier: PRP iteration timings are about 2.5% faster if you select an exponent that is a multiple of the FFT length.

I think I remember something about sieving being fatser when the binary representation of n contains mostly "0" bits. Are there any numbers which meet both criteria?

[Ken_g6]

Now you can thank PG for Sieving and for finding a twin prime!

Originally Posted by Puzzle-Peter

I think I remember something about sieving being fatser when the binary representation of n contains mostly "0" bits. Are there any numbers which meet both criteria?

I've been working on some optimizations to tpsieve at the end of this thread (and here), and with the hybrid Montgomery method, the first two bits should be 1, the next 4 don't matter so much (but more 1's is better), and then the rest of the bits should be 0. The closest number to 999999 that fits these criteria, for instance, is 999424. I don't know what the FFT length is, but I imagine more 0's is better there too.

[Mini-Geek]

Congrats to all involved for the new twin primes!

What are the primes?

[MooooMoo]

Originally Posted by Mini-Geek

What are the primes?

We'll reveal them within the next few days, once the discoverers have been contacted and given a chance to respond.

For now, I'll only say that the twin prime is 100355 digits long and has a k value between 65G and 67G.

[Flatlander]

You could be giving away more than you think.

[cipher]

Congrats to everyone who participated/helped out with the Twin project. Thanks PG & TP project. The question is Now what?

Now what do we do Mooomooo ? Give us some direction.

[MooooMoo]

I've looked over a previous poll: http://www.mersenneforum.org/showthread.php?t=6974 and some ideas on this thread. Based on that, I've made a preliminary decision to sieve the range 1<k<10M for 480000<n<500000. The chance of finding a twin in that range is slightly over 90%, and any non-twin primes found in that range will make it to the top 5000 list (their rank will be 3600th largest as of today).

Does anyone have any suggestions or comments? If there aren't any major objections, we'll start the sieve early next week.

[MooooMoo]

Originally Posted by geoff

tpsieve can sieve a range of n, that is a range of k,n pairs with k odd and k0 <= k <= k1 and n0 <= n <= n1.

However you need to start the sieve with NewPGen sieving each n separately until p > k1 at least, and then merge the resulting files for input to tpsieve (Just delete the header from all except the first file and then concanenate the results).

I can't seem to get this to work. I used a sample file called 0103, and it looks like this:

1000000000:T:0:2:3
18459 480001
59925 480001
94509 480001
9225 480002
36513 480002
14319 480003
16761 480003
67749 480003
98331 480003

The command I used was: tpsieve -i 0103 -p 1e9 -P 2e9 ,
but tpfactors only has:

1164933401 | 94509*2^480001+1
1231256737 | 18459*2^480001+1
1482274247 | 59925*2^480001-1

None of the n=480002 or n=480003 factors show up (1521564647 | 9225*2^480002+1, 1243389379 | 67749*2^480003-1, and some others were missing). Does anyone know what happened, and how to fix it?

[cipher]

Mooomooo i like your idea. I am onboard. Also referring to geoff previous post. We would have to do individual n's using newpgen and then combine them see if that fixes the problem? and then we can continue with tpsieve.
thanks
cipher

[mdettweiler]

Originally Posted by cipher

Mooomooo i like your idea. I am onboard. Also referring to geoff previous post. We would have to do individual n's using newpgen and then combine them see if that fixes the problem? and then we can continue with tpsieve.
thanks
cipher

That would essentially mimic the result of what MooMoo just did manually above. AFAIR, you have to either sieve each n separately using tpsieve or NewPGen (for tpsieve you can use a batch file, and for NewPGen you can use the increment counter), or the whole batch together with sr2sieve.