mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Riesel Prime Search (https://www.mersenneforum.org/forumdisplay.php?f=59)
-   -   Choose your own K and work on finding a top-5000 prime! (https://www.mersenneforum.org/showthread.php?t=4963)

 Kosmaj 2005-12-05 09:48

Selecting your k is the easiest part, just get an available one from the first post in the thread. Then you will have to sieve the candidates using [URL=http://primes.utm.edu/programs/NewPGen/]NewPGen[/URL], it's a free software, get it from the link.

Suppose your k=165, create a new folder and name it "165", unzip NewPGen in that folder and start it, than in NewPGen:
1) Select "Create a new file"
2) Enter the name of the output file, for example "165.txt"
3) Go to "Type", select "k.b^n-1 with k fixed" by scrolling it down
4) Set base=2, k=165 (your k), nmin=210000, nmax=400000.
5) Check "Verify results"
6) Select "Options", check "Log the numbers removed", and also check "Use two save files".
7) Click "Start" and let it run for 72 hours or until the sieving bound is 500 billion or more (displayed at the bottom of the NewPGen window).

Then stop NewPGen and feed the file from Step (2) above to LLR. If you have a variety of machines then run NewPGen on Athlon (or Duron, Opteron, or Pentium-3) and run LLR on Pentium-4.

The same applies to k*2^n-1, k<300 where you asked the same question but I couldn't find time to reply. Over there the bounds in Step (4) will be different since many k's are already checked.

 ValerieVonck 2005-12-05 09:57

Reserving 3250995

I will check from 10k => 300K.
My upper sieve limit will be 500G.

Regards,

 Kosmaj 2005-12-05 11:40

The sieving bound depends a lot on nmin and nmax (and in general case on [I]k[/I], but these are all high-weight ones, so it doesn't matter). If you go from nmin=10k you can stop sieving early, say at 20G, divide the NewPGen file into two parts, say to n=50k and the rest and then start LLR on the 10-50k range while keep on sieving 50-300k.

Also note that currently the "water line" to enter Top-5000 is at about n=207k, and if you go from n=10k by the time you reach 207k it will be already at 210k, that's why I suggested 210k above.

 ValerieVonck 2005-12-05 17:14

[QUOTE=Kosmaj]The sieving bound depends a lot on nmin and nmax (and in general case on [I]k[/I], but these are all high-weight ones, so it doesn't matter). If you go from nmin=10k you can stop sieving early, say at 20G, divide the NewPGen file into two parts, say to n=50k and the rest and then start LLR on the 10-50k range while keep on sieving 50-300k.

Also note that currently the "water line" to enter Top-5000 is at about n=207k, and if you go from n=10k by the time you reach 207k it will be already at 210k, that's why I suggested 210k above.[/QUOTE]

Thank you for claryfing things :bow:

 Kosmaj 2005-12-05 22:29

[QUOTE=BotXXX]k=8288233

LLR'ed
2000 < n < 491923

prime
8288233*2^148699-1 (but to small for top 5000 list)[/QUOTE]Thanks, I just updated the [URL=http://www.15k.org/lowweight.htm]Stats Low Weight[/URL]. But I think you should have sieved more, I just tried and removed next 23 candidates at the rate of "1 n every 52 sec."

 Jwb52z 2005-12-05 23:53

Kosmaj, there's some kind of problem with the link in your post to the Low Weight Stats page. It won't load properly.

 lsoule 2005-12-05 23:56

I've fixed the link - the .htm suffix was mixed up.

 Templus 2005-12-06 20:30

I sieved the file provided by BotXXX a bit more until 30bn. Anyone who wants this sieve file can mail or PM me.

k=10813783 has been tested until n=710000, 200 tests left until n=1M

 NiceMedTexMD 2005-12-07 01:16

my reservation...

wish me luck... :no:

 NiceMedTexMD 2005-12-07 01:29

when i'm running newpgen, I get a low rate... (1 n every xx seconds..) what kind of a rate should I get?

Mike.. :shock:

 lsoule 2005-12-07 01:44

It should start high but then drop off pretty quickly. Once you've sieved
up to 100M, the rate will probably be 1 per every second or two. Past
that, the rate will keep dropping. You can switch over to LLR when the
rate is roughly equal to what it takes to run LLR on one candidate (or
probably sieve past this since you'll start LLR at the lowest n and work
upwards where LLR will be slower).

All times are UTC. The time now is 17:12.