Sieving discussion thread

[Cybertronic]

I will upload a new file. Unzipped is it 714 MB. Name: twin333333-017.txt (zip: 123 MB)

NewPgen works fine with it.

First line:
"35909316 k's remaining. p=1175000396142961 divides k=48988743009"

[pacionet]

Originally Posted by MooooMoo

Pacionet, I can sieve the rest of the range (50G-207G), if your computer doesn't have enough RAM to sieve the whole 207G at once. It's a lot more efficient to sieve the whole range at once, than to split it into 50G ranges of k and then sieve each k range individually.

Please let me know if you want to sieve the whole range (1-207G) at once, or sieve 1-50G and let me sieve the rest (50G-207G). It's your choice, but the first option is more efficient.

OK, I'll sieve 0-50G because I have not enough RAM.
Do you have enough RAM to sieve 50-207 at once ? :surprised

[MooMoo2]

Originally Posted by pacionet

OK, I'll sieve 0-50G because I have not enough RAM.
Do you have enough RAM to sieve 50-207 at once ? :surprised

I have (barely) enough for sieving 50-207 at once. Sieving 1-207G at once is too much for my computer to handle, though

[Skligmund]

What kind of RAM requirements for that? Just curious.

[MooMoo2]

Originally Posted by Skligmund

What kind of RAM requirements for that? Just curious.

Around 800-900MB of RAM for the early parts (~2T), but only ~570MB is needed once the sieving depth is quite large (500T or more)

[jmblazek]

I recommend sieving the entire range at once, either 0G-207G or 0G-250G. Maybe someone can sieve it until RAM requirements lower and then turn it over.

OR, if splitting the file is necessary, then I suggest 0G-90G and 90G-207G or 250G. The faster computer would get the 0G-90G and the slower computer would get the other range. Reasoning...a twin is expected every 90G at n=500000...so if the dat file is going to be split, might as well sieve deeper into 0G-90G.

By the time 90G is LLR'd, the other range will be sufficiently sieved.

Also, 1G RAM can be as low as $80...so for a $160 investment 2G will adequately cover the RAM requirements.

[MooMoo2]

Originally Posted by jmblazek

Also, 1G RAM can be as low as $80...so for a $160 investment 2G will adequately cover the RAM requirements.

Only until the next n after 500,000. Beyond that, you'll probably need round 4G of RAM

Anyway, I've started sieving 50G-208G. My progress is at 26T. Unless something unexpected happens, it'll be close to (or at) the optimal sieving depth after n=333333 is complete, and after 1-50G is done for n=500000.

[Cybertronic]

With special methods is the 714MB file just 64.4MB large. Is it positiv ?

[gribozavr]

I achieved the same compression ratio.
Cybertronic, what methods did you use?

[Cybertronic]

I reduse in every line "333333".
Example: we have k = 3,9,13,21,45....
The file have all differences from number to number
k= 3,(+6),(+4),(+8),(+24) and so on
The reduce file unzipped is 160 MB.
Zipped is 64 MB.

To get the original output file, is a small "decode" program helpful.
This program create every line
k:
3 333333
9 333333 (3+6)
13 333333 (3+6+4)
....
for NewPgen

best

[gribozavr]

Biewma noted that the differences are divisible by 6. Also storing them in binary format helps.

Originally Posted by biewma

I recommend this format:
Header:TIPS1:<n>:<sieved_up_to_p>:<k's_left>
<first k>
Then differences divided by 6. in 2 Bytes.
(gives a range from 1 to 65535. Value 0 means add 65536 to next value). 0 0 10 gives for example (65535+65535+10)*6= a gap of 786540

What I have:
709,877,711 latest_sieve_file
71,384,246 latest_sieve_file.cpr
55,476,554 latest_sieve_file.cpr.zip