Page 2 of 4 FirstFirst 1234 LastLast
Results 41 to 80 of 142

Thread: Sieving

  1. #41
    Originally posted by Samidoost
    Their most successful application was to narrow the search for Sierpinski and Riesel problems. Currently there is remained 1 candidate for Sierpinski and 11 for Riesel.

    For the dual Riesel search you can see http://sierpinski.insider.com/riesel
    That URL is incorrect

  2. #42
    > That URL is incorrect

    Now it is corrected.

    Payam

  3. #43
    Originally posted by ceselb
    Cool. It'll take ~2 weeks to sieve this chunk for me. Is there any significant gain if you remove eliminated N's for the other peoples ranges as you go?
    Removing the numbers between 1 and 2M will save 40% of the time (i you haven't already done that.)

    I realized i have a save file at home which has all the candidates left when sieved to almost 300G. I removed a couple of more numbers already, but that doesn't really speed up things.
    Attached Files Attached Files

  4. #44
    Sander

    May I request the list of removed numbers? Your attached file does not contain them.

    Payam

  5. #45
    I didn't wanna make things too messy, so i just copy / pasted the removed numebers and sent them to your e-mail account.

  6. #46
    Thanks, Sander

    I will include them in http://sierpinski.insider.com/4847

    Payam

  7. #47
    One more question,

    is there any possibility to tell how many values you expect to find while sieving in a gieven range?

    I ask because i have sieved now from 600G to 612G and have no hits yet.
    Which looks not very effective in the first place.

    May be i find something in the next days.

    Lars

  8. #48
    Dunno, but all i know is that there can be a big difference in how many numbers you remove in a given range.

    I had a couple of times that no nimber was removed in over 6G, but the closest numbers removed had a prime factor less then 4M away.

    This is the breakdown of numbers removed in a 10G range (starting with 210-220G)

    2-6-4-8-6-4-5-3-9

    As you can see, a big difference.

    And the higher you sieve, the less candidates you remove

  9. #49
    Moderator ceselb's Avatar
    Join Date
    Jun 2002
    Location
    Linkoping, Sweden
    Posts
    224
    I wouldn't worry too much. I'm at 515G and haven't found any either. They'll come eventually.
    As I undersdtand it. these values will take much longer to process than the current ones, so a day or two apart is still worthwhile.

  10. #50
    I'm currently at 426G, and have found 7 numbers so far.

  11. #51
    I should have started complaining a little bit earlier.

    Only two hours after my mail i had the first four numbers removed.


    Lars

  12. #52
    Moderator ceselb's Avatar
    Join Date
    Jun 2002
    Location
    Linkoping, Sweden
    Posts
    224
    Then it's only me without any removed then... (I'm at 524.6B if anyone wonders)

    Nice to see some progress elsewere though

  13. #53
    On the primnubers mailing list Phil C. wrote:

    My choice of 3M as my sieve range was a curious one. As the sieving algorithm has roughly O(n^(1/2)) behaviour, where n is the range of the exponents (500000..3000000 would be 2500000), I have several criteria to balance:

    1) n not too large,
    a) as it's slower
    b) as several k would crack early, everything beyond that would be 'waste'
    c) as sieving next year is quicker than sieving this year! (~Moore's law)

    2) n not too small,
    a) as then the overhead of sieving vs. testing would increase
    b) as O(n^(1/2)) still favours large n, being better than linear

    If fewer than half of them have split by 3M then I'd say I got the balance
    right. If more than 8, I'd say we sieved too large a zone.

    For the future, the ones that reach n=2M probably should have sieving in a
    band of about 3M-20M started, with the same reasons as above.

    Just gut feel.
    Louie, Phil M,

    Any news about a customized newpgen which can do several K's at the same time?

  14. #54
    On the primnubers mailing list Phil C. wrote:

    My choice of 3M as my sieve range was a curious one. As the sieving algorithm has roughly O(n^(1/2)) behaviour, where n is the range of the exponents (500000..3000000 would be 2500000), I have several criteria to balance:

    1) n not too large,
    a) as it's slower
    b) as several k would crack early, everything beyond that would be 'waste'
    c) as sieving next year is quicker than sieving this year! (~Moore's law)

    2) n not too small,
    a) as then the overhead of sieving vs. testing would increase
    b) as O(n^(1/2)) still favours large n, being better than linear

    If fewer than half of them have split by 3M then I'd say I got the balance
    right. If more than 8, I'd say we sieved too large a zone.

    For the future, the ones that reach n=2M probably should have sieving in a
    band of about 3M-20M started, with the same reasons as above.

    Just gut feel.
    Louie, Phil M,

    Any news about a customized newpgen which can do several K's at the same time?

  15. #55
    Paul Jobling sent me the code required to do the multiple k sieve this morning. I am busy rewriting the SB client right now but I should find time soon to also write the threading code I want for the siever.

    I'll let you guys know how it performs when it's done. Based on Paul's email, it looks as though there may be quite a significant performance gain. We shall see.

    -Louie

  16. #56

    900G

    Hi,

    Last week, I started to work on 900G. Unfortunately I wasn't able to tell you about it since apparently my ISP's spam filters really don't like this forum's activation emails. After numerous attempts, I finally gave up and changed the address to a junk email account I have but never check...but I got the email.

    Anyway, since I was having problems getting it working, I started at 900G to reduce the chance that someone else would snatch up the range in the meantime. With the winter break coming up, though, I'm going to have to stop it soon, so it doesn't look like it'll get much beyond 950G. Good news is that it's removed about 10 numbers. I'll send them to ya either tomorrow or this weekend.

    Greg

  17. #57
    Great, just send me both the output file and the newpgen.del file.
    Make sure you clearly state which range you sieved!

    Today i recieved a mail from Paul Jobling with a new (Beta) version of NewPgen 2.71.

    I did a quick test, and for numbers with a low weight (like the ones we are sieving) it gives a speed increase of about 20% on my P3.

    Since i don't have any e-mail addresses of the people who are sieving i can't send it. So if you want it, drop me an e-mail

  18. #58
    Hi,

    i will do 700G to 800G also if that is not blocked ba anyone else.


    Lars

    P.S. The range from 600G to 700G will be finished today with over 20 numbers removed.

  19. #59
    Range from 600G to 700G is finished. Results are on the way to Sander.

    Finaly 27 numbers are removed.

    Lars

  20. #60
    Thanks

    27 numbers in a range of 100G is on average one number per 3,7G.

    At a wild guess, my P3 would remove 1 candidate about every 15 hours. I can't remember how long that pc needs for a test with N between 2 and 3M, But i think it's at least 3 days, probably getting closer to 5 if N get high into the range. So sieving is still at least 6 times faster at this depth.

    All wild guesses, and probably different for faster machines, but the point is that it's still worth sieving deeper.

  21. #61
    Well, I'm at 480G, removed 22 numbers so far. Should finish sometime in the next 2 days (maybe 3 max).

  22. #62
    The performance gain when using NewPgen 2.71 Beta
    is much higher for me.

    With version 2.70 i could crunch 0.286G per hour
    With version 2.71beta this is 0.48G per hour.

    Both on a PIII 700MHz.

    Brilliant job done by Paul Jobling.

    Lars

  23. #63
    Has anyone taken the 800-900G range?

    If not, I will start in on that range.

  24. #64
    Senior Member eatmadustch's Avatar
    Join Date
    Nov 2002
    Location
    Switzerland
    Posts
    154
    just a question out of curiosity ... why does everyone want to sieve the lowest sirpinski candidate 4847, when there are 12 others?
    EatMaDust


    Stop Microsoft turning into Big Brother!
    http://www.againsttcpa.com

  25. #65
    As far as i know 800-900 isn't taken by anyone.

    200 - 300 SMH (Done)
    300 - 400 SMH
    400 - 500 Bellbox
    500 - 600 ceselb
    600 - 700 LTD (Done)
    700 - 800 LTD
    800 - 900 Bewolf
    900 - 9?? frmky

  26. #66
    Originally posted by eatmadustch
    just a question out of curiosity ... why does everyone want to sieve the lowest sirpinski candidate 4847, when there are 12 others?
    27653 is already tested to 3M (only 5 numbers pending)
    33661 was tested to 1T earlier on (it was the second K the project started testing on.

    And the other 10 currentely are tested by louie

  27. #67
    why does everyone want to sieve the lowest sirpinski candidate 4847, when there are 12 others?
    Thats because all the other K are sieved by Louie.

    So only 4847 is our playground.

    Lars

    Edit:

    Hi Sander you have been a little bit faster with the reply.

  28. #68
    The 900G-1T range is finished. 22 numbers were removed, and the files have been emailed.

    Greg

  29. #69
    Paul Jobling did an great job speeding up NewPGen.

    v2.70=0.437G/hour
    v2.71b=1.12G/hour
    On a 1.4Ghz Tualatin Celeron @ 1.6Ghz

    Thats an incredible 156% speed increase

  30. #70
    I'm all done with 400-500 so I'll be emailing the output and deleted numbers shortly. Removed about 31 numbers.

  31. #71
    Hi

    I have started a new thread about k=28433 sieve. New results from Joseph Arthur up to 318G.

    Payam

  32. #72
    Range from 700G to 800G is also ready.
    Removed 19 numbers.

    Lars

  33. #73
    Just now i merged all the files that were sent to me. Until now, sieving has been done from 208G - 500G, from 600G - 800G and 900G - 1000G (so only 500-600 and 800 - 900 are missing).

    We started with 4221 candidates left above 2M (including a few Payam already proved composite). The *.del file now contains 202 numbers, of which 3 are double, so 199 candidates are removed and we are left with 4022 candidates.

    A break down of how many numbers were elliminated:
    • 208-300 49
      300-400 54
      400-500 32
      500-600
      600-700 27
      700-800 19
      800-900
      900-1000 22


    so we can expect about 50 more numbers in the two missing ranges.

  34. #74
    I'm about half done with the 800-900 range with 10 numbers removed so far. I should be done with the range by the end of the week.

    Are we going to stop at 1T? On average, sieving on my computer removes one number every 4.25hours. That is a lot faster than the SB client can test a number. Maybe it would be worthwhile to sieve a little farther.

  35. #75
    It's definately worth sieving further. Sieving is still a lot faster then prp testing, and even if sieving eliminates candidates at the same speed, you still won't loose anything.

    My P3(450) removed about 3 numbers a day around 350G, while it takes at least 3 (but could be 5, i don't have the logs anymore) days to do one proth test. Thats still 10 to 15 times faster!!

    My guess is that as long as we remove 4 or 5 numbers per 100G it's still faster than prp-ing. At least for my slow P3(450). I don't know how this is for a P4 or athlon.

    If anyone wants to take a range, please post it here. You don't necesarily have to take 100G, as long as you clearly tell what range you'll do/have done.

    I'll attach the merged file which has all candidates left after sieving to 1000G (except 500-600 and 800-900)

    One more thing, if i send you the beta of Newpgen v2.71, can you please check the date the exe was created. The program itself says 19-12-2002, but the executable should be of 20-12-2002.

    The 19-12-02 version contained a bug that misses some factors. I just want to make sure i sent everybody the correct version.
    Attached Files Attached Files

  36. #76
    my executable has a timestamp 20.12.2002 10:56 is that the correct version ??

    And by the way i will take 1000G to 1100G if nobody else has started this range.

    Lars

  37. #77
    Originally posted by ltd
    my executable has a timestamp 20.12.2002 10:56 is that the correct version ??

    And by the way i will take 1000G to 1100G if nobody else has started this range.

    Lars
    Thats the right one. I'm pretty sure i sent the right one to everybody, but i got a little confused by version numbers, different pc's , and two different e-mail accounts i used to recieve the beta. Just wanted to make sure.

    Okay, i'll continue at 1100G, it will take a while, but we're not in the 2M range with testing numbers yet.

    • 200 - 300 SMH (Done)
      300 - 400 SMH (Done)
      400 - 500 Bellbox (Done)
      500 - 600 ceselb
      600 - 700 LTD (Done)
      700 - 800 LTD (Done)
      800 - 900 Bewolf
      900 - 1000 frmky (Done)
      1000 - 1100 LTD
      1100 - 1200 SMH


    => EDIT:

    I got worried just now when i saw that there were only 4021 numbers to test, opposed to the 4022 i wrote earlier on.
    I was wrong, i once the lines in the output file and once included the first line, which is obviously no number to test, Payams file had a blank line in the end (and notepad just counts that as a line)

    So we started with 4220, found 199 different N's composite (and 3 duplicates) so there are 4021 numbers to test.

    MERRY CHRISTMAS ALL
    Last edited by smh; 12-24-2002 at 06:46 PM.

  38. #78
    Info from Bommer:

    He will do 1200G to 1300G.


    Lars

  39. #79
    Could you send me the new version of newPgen? I'd be happy to start sieving 1300G to 1400G.

  40. #80
    I've just sent v2.71 to you.

    Just to make things clear:

    • 200 - 300 SMH (Done)
      300 - 400 SMH (Done)
      400 - 500 Bellbox (Done)
      500 - 600 ceselb
      600 - 700 LTD (Done)
      700 - 800 LTD (Done)
      800 - 900 Bewolf
      900 - 1000 frmky (Done)
      1000 - 1100 LTD
      1100 - 1200 SMH
      1200 - 1300 Bommer
      1300 - 1400 Bellbox

Page 2 of 4 FirstFirst 1234 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •