Bases 6-32 reservations/statuses/primes

[michaf]

To give an idea of how large:
Starting riesel 31 without thinking resulted in a file of _shrug_ 14Gb

Originally Posted by gd_barnes

I was just thinking of that one myself but my resources are quite busy so fire away! The bases divisible by 3 generally drop the k's pretty fast so it shouldn't be too bad even with a high conjecture.

Recommendation: Run PFGW 4 times, 1 each for k == 0 mod 5, 1 mod 5, 2 mod 5, and 3 mod 5 up to n=5K-7K before eliminating multiples of 6 and sieving/LLRing. It'll make for much cleaner pfgw.out files that get pretty huge with a large conjecture such as this one has.


Gary

[mdettweiler]

Originally Posted by gd_barnes

Yes it is a power of 5, but the analysis is incorrect. k=25 is a perfect square. You should remove all n divisible by 2 because 5^2=25. If it was a perfect 5th power, i.e. k=3^5=243, then you would remove all n divisible by 5.

Analysis: let k=m^2 and n=2q...so m^2*30^(2q)-1 = (m*30^q-1) * (m*30^q+1) hence all even n are composite.


Gary

Oh, whoops. Thanks for correcting me. I wouldn't have wanted to accidentally remove the wrong n's!

The P3 is now happy sieving away on the k. I'm taking it up to p=600G (for an n-range of 25K-100K); is this adequate, or should I go further?

BTW, are there any other k's/bases that could especially use some presieved files? I'm thinking that this machine might be perfect for such hit-and-run sieving jobs, so I'm trying it out with this one Riesel Base 30 k.

[gd_barnes]

Originally Posted by Anonymous

Oh, whoops. Thanks for correcting me. I wouldn't have wanted to accidentally remove the wrong n's!

The P3 is now happy sieving away on the k. I'm taking it up to p=600G (for an n-range of 25K-100K); is this adequate, or should I go further?

BTW, are there any other k's/bases that could especially use some presieved files? I'm thinking that this machine might be perfect for such hit-and-run sieving jobs, so I'm trying it out with this one Riesel Base 30 k.

Without doing some testing, off the top of my head, I'd say that was a little low. Base 30 is very LARGE and sr1sieve is fast! Typically now, I'd suggest what Curtis does; to sieve entire thing, break off a lower piece, and sieve remainder but I'm sure you don't want to hassle with that with a borrowed machine. For this, just LLR a candidate at 70% of n-range, i.e. n=77.5K, and sieve until the removal rate equals the LLR time. That LLR time should be very long! With speedy sr1sieve, optimal may be P=800G-1T. P=400G-600G is good for lower bases for n=25K-100K, depending on weight and # of k's.

Late estimate without running an LLR test: I checked my results file from when I ran n=0-25K. 25*30^19391-1 LLR'd in 71.9 secs. n=19391 is about 1/4th of n=77500, which means n=77500 would LLR in 71.9 secs. x 4^2 = 1150 secs. on a Dell core duo 1.66 Ghz. When you cleaned your machine out, your timing was close to mine so you should sieve until the removal rate is around 1150 secs.

Edit: If you want to have a little fun with sieving a HUGE base that is a power of 2, try Sierp base 256 with 2 k's remaining. k=535 has been tested up to n=53.7K and k=831 has been tested up to n=12.5K. The equivalent base 2 is 8X as large. Use your judgment about what range to sieve. If sieving a very large n-range, consider sieving, breaking off, and then sieving some more.

One thing I want to do with this effort in the near future is get some k's to test that are top-5000 size for a base that is a power of 2 but that are not inordinately large (~n=400K-800K base 2 equivalent). We'll get that with the team drives when we get them all sieved to n=200K, but base 256 would be a way to get a jump start on that size of potential prime.



Gary

[gd_barnes]

Reserving k=1478 and 3620 for Riesel base 16. I'll double-check them up to n=65K and 75K respectively and then take them on up to n=100K.

I'll also double check Curtis's k=443 from n=25K-65K.

The test limits came previously from the Prime Search site converted from base 2.


Gary

[Xentar]

Just a short status update:
base = 18; k = 122; n ~ 140K, continuing.

Edit:
Ehh, I think, I forgot the most important detail: No primes yet :(

[geoff]

Testing is at n=8K, there are 95 candidate k left.

[Siemelink]

Hidiho,

I've finished my range on 4233*22^n+1. I won't be continuing with it.
Some more of my ranges will finish soon. In prepration I'll take
342*27^n+1
398*27^n+1
278*30^n+1
588*30^n+1
all from 25k to 100k.

Laters, Willem.

[gd_barnes]

Originally Posted by Siemelink

Hidiho,

I've finished my range on 4233*22^n+1. I won't be continuing with it.
Some more of my ranges will finish soon. In prepration I'll take
342*27^n+1
398*27^n+1
278*30^n+1
588*30^n+1
all from 25k to 100k.

Laters, Willem.

Willem,

Just to confirm. You completed 4233*22^n+1 to n=100K and you are running both 1611*22^n+1 and 5128*22^n+1 up to n=200K. Is that correct?


Thanks,
Gary

[Siemelink]

Originally Posted by gd_barnes

Willem,

Just to confirm. You completed 4233*22^n+1 to n=100K and you are running both 1611*22^n+1 and 5128*22^n+1 up to n=200K. Is that correct?


Thanks,
Gary

Still true. The k = 1611 is at 173,000, 6000 seconds per test. k = 5128 is up to 138,000.

Laters, Willem.

[sjtjung]

706 for base 27 is done to 50k. No primes found and I'm releasing it.



-Steven

[Siemelink]

Originally Posted by sjtjung

706 for base 27 is done to 50k. No primes found and I'm releasing it.
-Steven

Ok, I'll pick this one up and take it to 100k. I had sieved it before christmas before I noticed that you had reserved it.

Willem