Sieving. 300<k<=1001, n=600K-1M.

[mdettweiler]

Originally Posted by gd_barnes

Responded to. As I allude to in the response, we'll also need some sieving for 17-18 k's < 300.

We could have Lennart finish k=300-1001 up to P=25T (or whatever optimum is) and have you do the k's < 300 after Anon and I get it started.

Lennart, do you feel you'd have the available resources for sieving k=300-1001 for all of P=14T-25T over the next ~2 months?

Bruce, are you OK with sieving 17 or 18 k's < 300?. It should go 4-5 times as fast as with 300-350 k's. Anon and I would get it started and if you get tired of it or run short of resources, we could get it finished up.

Anon, what do you think of doing it that way?

One thing I think I'll do: The first 2 primes that I find for n>600K from these efforts, I'll split one with Lennart and one with Bruce. We greatly appreciate the huge contribution on sieving.


Gary

Okay, that sounds good. Bruce, what do you think of doing it this way? I'll be getting the sieve started within the next day or two, so if you're interested just give me a holler and I'll send you the sieve file.

[Brucifer]

Originally Posted by Anonymous

Okay, that sounds good. Bruce, what do you think of doing it this way? I'll be getting the sieve started within the next day or two, so if you're interested just give me a holler and I'll send you the sieve file.

That way is okay with me. That's different than what I had answered to in the PM. However given the circumstances, it's probably better to do it that way, and then later on if it looks like time is going to be an issue on the stuff Lennart is working on I can shift back over to that.

So send me the dat file and stuff whenever you are ready Anon.

[gd_barnes]

Originally Posted by em99010pepe

Lennar,

With n up to 1M we need to go to P=30T or 40T but I suppose we will only do that to the range 900k<n<1M.

Carlos

Optimum for sieving the entire k=300-1001 range to the same depth is P=~26T. This is based on LLRing an n=880K candidate and the sieve rate both on modern non-overclocked 64-bit machines. If you overclock your LLRing machines, the optimum would be LOWER so the depth is a good thing!

Likely within about the next 2 months, our individual-k drive will have k's sieved to P=26T. Further: All of k=300-1001 for n=600K-1M will be sieved to P=26T!

This should be very exciting, almost like starting the project new!


Gary

[gd_barnes]

Originally Posted by Brucifer

That way is okay with me. That's different than what I had answered to in the PM. However given the circumstances, it's probably better to do it that way, and then later on if it looks like time is going to be an issue on the stuff Lennart is working on I can shift back over to that.

So send me the dat file and stuff whenever you are ready Anon.

I'm going to suggest that we sieve the 18 k's < 300 to P=26T also. If you have the available resources to set that up and run, that would be great. Like I stated previously, I'm thinking that it should take about the same amount of CPU time as your sieving k=400-1001 to P=4.9T took.

This is likely over-sieving a bit for so few k's but I'd rather have it sieved to the same depth as our main range. Also, since we have people who like to sieve right now, we should utilize them while they have the available machines.


Gary

[em99010pepe]

What's the ETA to sieve 18 k's < 300 to P=26T? I am 4 days away to have 4 cores available to start LLRing those 18 k's...

[mdettweiler]

Originally Posted by em99010pepe

What's the ETA to sieve 18 k's < 300 to P=26T? I am 4 days away to have 4 cores available to start LLRing those 18 k's...

I'm not sure. Bruce, do you have a ballpark ETA for that?

Carlos, in the meantime, I've got k=289 already sieved for n=520K-600K so we can get it caught up to p=600K before we start on it with the rest of the k's. If you'd like, I can give you that to work on--my estimate is that it's about two weeks' worth of CPU time on a Core 2 Duo 2.2Ghz. (Or, if you'd rather, I can load it into a "private" LLRnet server at http://llrnet.ironbits.net--in fact I've already got two servers on there hosting personal efforst for me.)

[em99010pepe]

That k is already tested from 600k to 1M.
Right now I am cleaning my server so that is out of the question, thanks for the offer.

Carlos

[mdettweiler]

Originally Posted by em99010pepe

That k is already tested from 600k to 1M.
Right now I am cleaning my server so that is out of the question, thanks for the offer.

Carlos

Well, some of the validity of the results of that range is in question, since it was only publicly reported up to n=520K, so hence we're essentially doublechecking all or part of its range in this drive (just like we are doing, sort of, for numerous 300<k<1001 in drives #1 and #3).

[em99010pepe]

New versions available for sr1sieve and sr2sieve.