Sieve question

as long as we have a moderator i nthis forum I was wondering if one of you could offer your input on the 20-100 million sieve. I've started on the dat and I've already sieved 0-3G It is running quite slow but I am hoping that with some help from MKlasson we'll be able ot have one as fast as the original proth sieve soon. my only real desire for this project (20-100 million sieve) at the current time is that we get the range sieved enoug hso taht the prospect of doing prp tests in the range of a 10,000,000 digit prime would be possible. Not that i believe we are likely to find a prime there but if the option exists we could attract a number of users who would be attracted by the reward money that would come with the discovery. Understanding that many prime searchers are drawn to gimps because of their promise to give the finders of primes 5000 dollars and the reward of 100,000 dollars to the finder of a 10,000,000 digit monster. If we have a shot at contending for this goal the nwe could also offer a similar such enticement (on the condition that we are the finders of the prime). Its just an idea I've been tossing around in my dreams at night but anything that can boost our membership is a good thing right. As far as the sieving of the 20-100 million range I'll handle it up to 3T once i can get a efficient siever running. Right now i am luimited to a extremely slow pace because of having to use older sievers and having to deal with memory constraints on certain machines. I hope to getaround this soon enough though. I'm also hopping taht if we can get the 20-100 million sieve to function at a reasonable speed then perhaps we could add 1-20 million and continue from our current location i nthe regular sieve. This would of course give new life to the sieving effort which is predicted to die out in the next few months with the dropping factor density taht we've been experiencing. This increased range would give an incredible boost in efficiency that would make continueing the sieve well past 1000T extremely reasonable. Many people have said taht the discovery of another prime in the mean time would make the addition of the extra sieving range a waste but primes don't come with any regularity and it needs to be considered the incredible amount of effort that is saved by doing the sieves together instead of seperatly. Also many more tests would be saved in the current range (1-20million) if the additional range was added because there would now be reason to continue the sieve to much greater depths.

Quote:

---

Originally posted by Keroberts1
as long as we have a moderator i nthis forum I was wondering if one of you could offer your input on the 20-100 million sieve. I've started on the dat and I've already sieved 0-3G It is running quite slow but I am hoping that with some help from MKlasson we'll be able ot have one as fast as the original proth sieve soon. my only real desire for this project (20-100 million sieve) at the current time is that we get the range sieved enoug hso taht the prospect of doing prp tests in the range of a 10,000,000 digit prime would be possible. Not that i believe we are likely to find a prime there but if the option exists we could attract a number of users who would be attracted by the reward money that would come with the discovery. Understanding that many prime searchers are drawn to gimps because of their promise to give the finders of primes 5000 dollars and the reward of 100,000 dollars to the finder of a 10,000,000 digit monster. If we have a shot at contending for this goal the nwe could also offer a similar such enticement (on the condition that we are the finders of the prime). Its just an idea I've been tossing around in my dreams at night but anything that can boost our membership is a good thing right. As far as the sieving of the 20-100 million range I'll handle it up to 3T once i can get a efficient siever running. Right now i am luimited to a extremely slow pace because of having to use older sievers and having to deal with memory constraints on certain machines. I hope to getaround this soon enough though. I'm also hopping taht if we can get the 20-100 million sieve to function at a reasonable speed then perhaps we could add 1-20 million and continue from our current location i nthe regular sieve. This would of course give new life to the sieving effort which is predicted to die out in the next few months with the dropping factor density taht we've been experiencing. This increased range would give an incredible boost in efficiency that would make continueing the sieve well past 1000T extremely reasonable.

---

I moved this from the other location. I don't know much about the sieving, this is a question Louie or others should answer, not me.

sorry all i knew of your input was that you were an administrator. Hopefully they'll see it soon.

Keroberts, if your final aim is to get a good sieve dept at 10 million digit candidates, I'd recommend you to compile a compact range for n that produces 10 million+.

There's no need to start a slow 20m-100m sieve for this purpose.

I don't recall the exact n, but let's say it's 33,218,500. Then, simply make a dat file for nmin=33218500 and nmax=34218500.

This will get you enough number of 10 million digit numbers to test, and sieving would have been done relatively quicker.

But, to be honest, I am not sure if there will be some demand for that.

The most imporant issue is, as you know, our PRP client is roughly 4 times slower at equal sized n.

This would suggest a year to complete a 10 million digit PRP test on an average machine.

And, (I might be mistaken on this, but anyways) the odds of finding a prime (per test of equivalent size) might be lower.


PS: To my opinion, starting a sieve of larger n ranges is another issue. If we get an efficient siever (with less than, say 30%-40%, speed loss at 1m-100m w.r.t. 1m-20m), I guess most of the people here would like to switch earlier. But if not, I think Mike's point is right.

is there any use for factrange.txt?

Nothing significant that I can think of.

What about the low n k/n pairs in factrange.txt?

Well, all n<750K have been at least double checked, so a factor for 750K<n<800K would save a double check PRP test (not really a lot, but some hours on slow PCs anyway). For 800K<n<1M, it might save more, but that's not clear right now...

I just wrote a little program to exclude said factors from the factrange.txt:

http://www.mystwalker.de/lowNFinder.class

Just put it into the same directory as the text file and execute it via "java lowNFinder" (if some error message appears, try "java -cp . lowNFinder").

It outputs 3 text files:

- lowrange.txt: All factors 1000 < n < 1000000 (all low n factors)
- medrange.txt: All factors 750000 < n < 1000000 (those low n factors still helpful for some double check)
- DCrange.txt: All factors 750000 < n < 800000 (those low n factors still helpful for the current double check)

Quote:

---

Originally posted by Nuri
Nothing significant that I can think of.

---

one can put them into

Pos UserName Score (%) FacU (%) FacD n<3M 3M<n<20M n>20M - this column :p