Bases 6-32 reservations/statuses/primes

[gd_barnes]

Originally Posted by jasong

I'm still hoping for a sieve file.

Also, I've thought about it, and I've decided I'm willing to sieve for any base from 2-31. I'm thinking I might download the various sieve files and see what kinds of numbers I come up with.

Of course, I intend to start with my already reserved number. k=16734 base=4. :)

What kind of sieving resources do you have? There are 2 'factoring holes' in the Sierp base 16 drive for n=100K-200K up to P=1T. They are at P=772.8G-800G and 976.5G-1T. After the holes are filled in , we need to take it up to P=1.5T or so.

I'm thinking to get it to 1.5T would take about 10 CPU days but just getting it to about P=1.2T-1.25T would be helpful.

I need to remove k's found prime by the drive in the last few days but I could send them to you within about 6 hours or so. (4 AM GMT)

Let me know and I'll give you more specifics about what is needed.


Gary

[jasong]

Originally Posted by gd_barnes

What kind of sieving resources do you have? There are 2 'factoring holes' in the Sierp base 16 drive for n=100K-200K up to P=1T. They are at P=772.8G-800G and 976.5G-1T. After the holes are filled in , we need to take it up to P=1.5T or so.

I'm thinking to get it to 1.5T would take about 10 CPU days but just getting it to about P=1.2T-1.25T would be helpful.

I need to remove k's found prime by the drive in the last few days but I could send them to you within about 6 hours or so. (4 AM GMT)

Let me know and I'll give you more specifics about what is needed.


Gary

That'd be great(receiving the sieving files, I mean), I have to play catchup with some non-Mersenne forum stuff(I'm very forgetful, need to take better notes), but I ought to be able to run the sieving in the next few days, and be done in a week or so. 10 cpu days sounds pretty easy, considering I have a Core2Quad. :)

[gd_barnes]

Originally Posted by jasong

That'd be great(receiving the sieving files, I mean), I have to play catchup with some non-Mersenne forum stuff(I'm very forgetful, need to take better notes), but I ought to be able to run the sieving in the next few days, and be done in a week or so. 10 cpu days sounds pretty easy, considering I have a Core2Quad. :)

I just realized that I'll have a dual-core machine freed up later tonight. I'd like to fill in the 'factor holes' first so that it is cleanly sieved up to P=1T. That should take < 1 day. I'll then send the files to you for sieving to P=1.5T. Look for them sometime late Thurs./early Fri.

Are you familiar with how to run sr2sieve? If not, I should mention that it doesn't remove candidates, it only writes out factors. I'm fine if you want to run the sieve and send me the factors. I can then remove the candidates using srfile.


Gary

[tcadigan]

Originally Posted by tcadigan

...
sieve-sierp-base16-13438.txt
...

13438*2^395260+1 is prime! Time: 232.374 sec.

aka 13438*16^98815+1

results file attached. (file completed)

Gary can you send a pm with directions how to submit properly? I do believe it's large enough and I've never done it before...

reserving:
sieve-sierp-base28.txt

[gd_barnes]

Originally Posted by tcadigan

13438*2^395260+1 is prime! Time: 232.374 sec.

aka 13438*16^98815+1

results file attached. (file completed)

Gary can you send a pm with directions how to submit properly? I do believe it's large enough and I've never done it before...

reserving:
sieve-sierp-base28.txt

Wow, lightning strikes twice in two days! Congrats on your first top-5K prime!

And just in time before the file ran out. It saves me sieving time for n=100K-200K also. (I'm including all k's remaining in it; not just the 'team drive' k's.) Life is good!

I'm thinking it makes sense to publically post the instructions on submitting a top-5K prime. I'll do that in the 'report primes here' thread in a little while.

Edit: This one will be a little 'different' so to speak. You can choose to submit it as is but the top-5K site will 'normalize' it to 6719*2^395261+1. I would suggest submitting it in this normallized format. I prefer to do it that way because otherwise you get a cryptic message that is not an error but is hard to understand if you've never done it before. I'll send you a PM in addition to the public instructions.


Gary

[michaf]

No, I did not add it back (haven't even seen it coming up :>)

I'll add it to my files, and sieve/llr it as far as the rest

Originally Posted by gd_barnes

Micha,

Did you add back MOB k=17496 to Sierp base 24? I had tested it to n=6.5K with no prime and had added one to your remaining k's from before. I had assumed that you had previously removed it per the prior project description. So this would now make 170 k's remaining unless you found a prime for it.


Gary

[Xentar]

Sierp b17: LLR around n~90k for all three k. I reserved till 100,000, but I think I will continue. Will start sieving soon, but have a question again.
What's more effective:
- Sieve all three k together, or alone? Alone, I can use sr1sieve, which is much faster..
- Sieve n = 100,000 - 200,000, or should I already sieve a bigger range (for example 100,000 - 1,000,000)?

Sierp b18: LLR around n~150k, still one k remaining, and still no prime :(

[Siemelink]

342*27^36291+1 is prime.

found PRP by LLR, proven with pfgw. I won't continue that range obviously.
Has anyone ever found a PRP that turned out not to be prime after all?

Willem.

[michaf]

Usually, taking a bigger range doesn't affect the speed too much.
Then again, whenever you find a prime, the extra time is 'wasted'.

I'd say that after n=0-100k, a good choice would be n=100-500k.

Best would be to test-sieve a few ranges and see how it affects the speed

[gd_barnes]

Originally Posted by Siemelink

342*27^36291+1 is prime.

found PRP by LLR, proven with pfgw. I won't continue that range obviously.
Has anyone ever found a PRP that turned out not to be prime after all?

Willem.

Excellent! Only one k to go on Sierp base 27. If you knock that one out, it could be our first proof of a conjecture!

I know I haven't found a PRP to be not prime.

Edit: I think there is some smaller ones listed on the top-5000 site. Supposedly the larger the PRP, the less chance it has of being not prime. I heard someone say that there's a much better chance of a hardware error causing an incorrect prime than a PRP being found not prime.


Gary

[gd_barnes]

Originally Posted by michaf

Usually, taking a bigger range doesn't affect the speed too much.
Then again, whenever you find a prime, the extra time is 'wasted'.

I'd say that after n=0-100k, a good choice would be n=100-500k.

Best would be to test-sieve a few ranges and see how it affects the speed

Agreed on the range to sieve. The speed varies by the square root of the n-range so n=100K-500K should sieve about 50% slower [1-1/sqrt(100K/400K)] than n=100K-200K. If you don't want that kind of slowdown, you could do n=100K-300K, which should sieve about 29% slower [1-1/sqrt(100K/200K)]. The main thing here is that on these low-weight k's, the chance of finding a prime in any n=2X range, i.e. n=100K-200K, is well under 50% so the bigger range is better. Clearly you'll want to break off pieces of the range and LLR them as you go while continuing to sieve the higher ranges.

About how to sieve: For 2 k's, you would generally use 2 instances of sr1sieve. But for 3 to ~50-100 k's, use sr2sieve on all of them. >~50-100 k's use srsieve for all sieving. There may be exception situations on 2 or 3 k's since that's kind of the dividing line on what to use. But sr1sieve running 1 k would have to be 3X as fast as sr2sieve running 3 k's, which seems unlikely in most situations. You might test it out to be sure and post back here what you find out.


Gary