Sieving for CRUS

[VBCurtis]

Originally Posted by sweety439

Hi,
I am a newcomer, I want to sieve these bases to BOINC, how to sieve?

How much work (CPU time) are you interested in doing?

srsieve is used to create the file and do initial sieving. Once you reach 1e9, use sr2sieve for most bases (a base with 1-3 k's remaining may be faster with sr1sieve on each k individually rather than sr2 on all of them at once).

The flags needed for invocation of the programs can be learned via "srsieve -h" etc, or surely there's a thread detailing exact invocation here in this forum. sr2sieve can be multi-threaded with -t 4 flag (for 4 threads); this allows modern machinery to run a sieve pretty quickly. For example, I think I spent about a month on an older 4-core Intel chip to sieve R327 earlier this fall to 80T or so, sufficient for testing the first part of the file.

In general, it's not very efficient to sieve ranges smaller than a factor of two (say, 2M-3M).

[Soapbox not necessarily shared by the regulars in this forum:]
200k-500k is more efficient than 200-400k, but a too-large range (say 200k to 5M) is a waste of work because any prime found removes that k from the testing pool, "wasting" all the sieve computations for that k above the prime. I personally choose the range of exponents such that 1/3rd or less of the k's are expected to prime when testing the file; a factor of 3-5 usually fits this concept (say, 200k to 1M or 300k to 1M). It's not a big deal, but the way I see it a 10% gain in efficiency is equivalent to a free 10% overclock, and I go to great lengths to achieve those!
[/Soapbox]

[VBCurtis]

Originally Posted by MisterBitcoin

By the way, sieving depht for S17. I suggest p=90T or just the time to eliminate 1 canidate. (60% Rule.) ?
N-Value is going up to 5M.

If the sieve range is around a factor of two (say, 200k to 400k or 500k to 1M), the rule of thumb is to sieve until the average time per factor is equal to the time to primality-test a candidate 60% of the way into the file (from 200k to 400k, that's 320k exponent: 60% of (400-200) + 200). If you knew the entire file would be primality tested, the rule of thumb is 70%; but fewer of the candidates at the end of the file will be tested because we expect to find some primes in the lower ranges, so we shift the rule of thumb lower (60% is widely used around here).

You don't need to run a full primality test of such a candidate to get this timing; you can run a few minutes of the test for iteration timings and multiply by the number of iterations needed.

[lalera]

Originally Posted by sweety439

Hi,
I am a newcomer, I want to sieve these bases to BOINC, how to sieve?

S9 2M-3M
S30 500K-1M
S46 100K-200K
S48 100K-200K
S212 all k's to n=1M
S214 200K-1M
S218 200K-1M
S228 200K-500K
S257 600K-1M
S263 350K-1M
S264 200K-400K
S294 200K-400K
S320 200K-400K
S332 k=4 to n=1M
S334 k=4 to n=1M
S426 400K-1M
S461 250K-1M
S476 200K-400K
S480 200K-400K
S497 200K-400K
S579 200K-400K
S836 200K-1M
R6 5M-10M
R30 k=659 and k=1024 to n=1M
R45 250K-400K
R55 250K-400K
R60 k=36 to n=1M
R102 200K-1M
R157 100K-200K
R159 600K-1M
R213 200K-1M
R214 500K-1M
R218 200K-1M
R221 400K-1M
R233 400K-1M
R234 400K-1M
R236 400K-1M
R253 500K-1M
R275 400K-1M
R328 900K-1M
R398 200K-1M
R425 400K-1M
R581 200K-1M
R650 200K-1M
R812 200K-1M
R968 250K-1M
R1019 400K-1M

hi,
in the case you are using linux x64
an example for S476, k=28, n=200k to 400k
you could start with
./srsieve-x86_64-linux --newpgen --nmin 200000 --nmax 400000 --pmax 20e9 28*476^n+1
this will take about 30 minutes
take the output file t16_b476_k28.npg
as an input file for sr1sieve
./sr1sieve-x86_64-linux -i t16_b476_k28.npg -o b476k28-1.txt -P40e12 -t4
this will take about 6 days

[gd_barnes]

Originally Posted by sweety439

Hi,
I am a newcomer, I want to sieve these bases to BOINC, how to sieve?

S9 2M-3M
S30 500K-1M
S46 100K-200K
S48 100K-200K
<snip lots of bases>

I had missed the post initially Please reserve the sieving for only 4-5 bases at a time as a starting point. The effort required for each base is likely larger than what you are initially anticipating. It is difficult for the project to manage such large reservations and you may find that your initial enthusiasm wears off after doing many of them. After you've done 4-5 of them then just reserve a few more. Thanks.

[sweety439]

Thanks.

Now, I only want to sieve 3 bases:

S30 to n=1M
S461 to n=1M
R1019 to n=1M

[MisterBitcoin]

R549 sieved to p=20T, file attached.
Next one (R533) will be completed in 24 hours.

By the way:
Reserving S1021

[MisterBitcoin]

R533 is completed to p=20T.

[lalera]

Originally Posted by lalera

hi,
here is a sieve-file for R565
n= 100k to 200k
sieved to 44T

to gd_barnes:
may i remind you that R565 is already done
and a question for R742:
on the website and in the sieve-file there are 78 k´s
but you wrote 79 k´s
what is right?

[gd_barnes]

Originally Posted by lalera

to gd_barnes:
may i remind you that R565 is already done
and a question for R742:
on the website and in the sieve-file there are 78 k´s
but you wrote 79 k´s
what is right?

Both issues fixed. Thanks for pointing them out. There are only 78 k's remaining for R742 so the sieve file is right and the pages have been corrected. When people post sieve files I do not closely inspect them so I welcome anyone to closely inspect the k's being sieved in the file. All that I generally do is take a cursory glance at the header to make sure the base and sieve depth are what the person posted.

[MisterBitcoin]

Originally Posted by MisterBitcoin

R523 is finished to p=20T.

Next one (R549) will be completed in 2 days.

Completed aswell, information isn´t updated sofar.

[gd_barnes]

Originally Posted by MisterBitcoin

Completed aswell, information isn´t updated sofar.

Fixed.