status on k=157 and k=161
 

k=157 and k=161 are done through n=200000 now.

157*2^n-1 is prime for: n=1, 5, 33, 97, 121, 125, 305, 445, 473, 513, 1581, 3077, 3517, 6345, 8797, 15265, 16641, 32185, 32581, 43905, 77025, 84377, 202345

161*2^n-1 is prime for: n=2, 6, 18, 22, 34, 54, 98, 122, 146, 222, 306, 422, 654, 682, 862, 1010, 1346, 1582, 7890, 9714, 12498, 21742, 27810, 34798, 74382, 97746

-- Thomas L10

[QUOTE=Keller]How deep do u think should I sieve the range ?[/QUOTE]

[QUOTE=Kosmaj]Well, if you plan to stop at n=200k then to 50bn, for example.[/QUOTE]

Well, it depends not only on the upper bound, but also on the weight of your k. There are k for which 30bn would be sufficient and other k for which you need to go up to 300bn. Since k=75 is divisable by 3 and 5, it is one of the "heavier" ones. The Nash weight is 3181. Therefore I suggest at least 100bn, if you plan to go to n=200000. For nmax=250000 you should think of 250bn or more...

The general rule is to sieve as long as the sieve eliminates faster than LLR. Since the time for a LLR test depends roughly on n^2, it isn't quite easy to find the right time to stop the sieve.

First of all you should choose an apropriate lower and a upper bound, e.g. nmin=2 and nmax=200000. Think of the time you want to spend on your specific k. So don't take nmax too high!
Then run a LLR test on k=<your k> and n=nmax and another LLR test on k=<your k> and n=nmin. If nmin<1000 you may skip the latter and set t(nmin) = 0. In principle you don't need to do a full test on n=nmax, since it is sufficient to get the time per iteration. From that you can extrapolate the time for a full test on n=nmax.

Now there are different approaches to get the apropriate sieve depth from the times t(nmin) and t(nmax). One rule of thumb is to sieve until the elimination rate drops below one candidate in (tmax+tmin)/2. Others take (tmax+tmin)/1.67, since in average you spend much more time on the higher n. But if you take the lower ranges out of the sieve for LLR testing before reaching the "optimal" sieve depth and continue the sieve only for the higher ranges, then things are different again.

The more k's you have tested the more you will get a feeling for the "optimal" sieve depth ...

-- Thomas

And here comes another one for k=161:

161*2^205370-1 is prime!

Seems that setting nmax=200000 is a "bit" too low :whistle:

-- Thomas L10

To cite Lee Stephens, founder of the RieselSieve project: "Primes came in bunches":

157*2^230777-1 is prime!
157*2^234401-1 is prime!

-- Thomas L10 :smile:

261
 

Joss

Think I'll take on k=261 up to n=200K. I am at 475000 on k=15 with no further primes.

Regards

Robert Smith

k=79
 

79*2^214453-1 is prime! :smile:

Other small primes for n = 65977, 82081, 92517.
Checked 20000-100000 and 163000-236000.

Searching further k=7, 13, and 79.

I agree than nmax=200000 is too small, the client is fast and I suggest we increase nmax at least to 250000. Also, to report a prime to Top-5000 nmin=163000, so the 163000-200000 window is too narrow.

More k's for LLRnet
I'm starting to sieve the following K's for LLRnet on riesel.15k.org! (N=250 to 250,000)

195, 197, 199, 201, 203, 205, 207, 209, 211, 213, 215, 217, and 219!

You can stop sieving n<=10000 for all k's, I'll check those using Proth.exe. it's much faster and we don't have to waste time passing over those small n's over the network. Will post found primes below, it will be done in 20-30 minutes. :smile:

Edit: I checked all 13 k's in the 250-10000 range. All found primes match exactly those on Wilfrid Keller's list. According to the list all k's are already checked to 32000, so we're just double checking those. Proth.exe als found that:[CODE]
219*2^657-1 = 219*(2^3)^219 - 1, Gen Woodall base 8
201*2^2413 - 1 = 402*(2^6)^402 - 1, Gen Woodall base 64
213*2^726 - 1 and 213*2^726 + 1 are twins.
[/CODE]

[QUOTE=Kosmaj]You can stop sieving n<=10000 for all k's, I'll check those using Proth.exe. it's much faster and we don't have to waste time passing over those small n's over the network. Will post found primes below, it will be done in 20-30 minutes. :smile:[/QUOTE]

We shoud test all n under 25000 this way. The server doesn't work well with small numbers.

Citrix
:cool: :cool: :cool:

[QUOTE=Kosmaj]You can stop sieving n<=10000 for all k's, I'll check those using Proth.exe. it's much faster and we don't have to waste time passing over those small n's over the network. Will post found primes below, it will be done in 20-30 minutes. :smile:[/QUOTE]

I have done some up to (20k). There's only a few left

[QUOTE=Citrix]
We shoud test all n under 25000 this way. The server doesn't work well with small numbers.[/QUOTE]

That sounds good.

When Mark gets the server in top shape. There will be more working on LLRnet

Joss

Then let's test them all to 32k, thus completing all double-checks off the network. Just tell us the status and let's divide the work, either by k or by n. I tested them all to 10k (see above).

BTW, I'd also like to reserve [b]k=77[/b] to test by myself to n=250k.