Loading ...
Sorry, an error occurred while loading the content.

17746Re: [PrimeNumbers] Re: symmetrical primes

Expand Messages
  • Phil Carmody
    Mar 12, 2006
    • 0 Attachment
      [note - I've changed my primenumbers settings from daily digest to individual
      mails, so do not need an expedited copy sent directly to me any more.]

      --- Alan McFarlane <alan.mcfarlane@...> wrote:
      > I'm just in the process of writing an optimized version (ANSI C to start

      If your C is not fast enough and you're tempted to go into assembler, then
      you're not writing fast enough C! (Not always true, of course, but I am
      notoriously pro-C.)

      > with), but in the interim, why don't we allocate blocks of data to work on.
      > Say a block size of 2^40? Seems a reasonable size to me - it's fairly
      > easy for most machines to do, but is still meaty enough to be interesting.

      That does make some sense. I'll look at optimising my code too, and decide how
      much I want to bite off. I'd normally have my machines running PRP-ing for Les
      GeneFermiers, but can certainly put aside a few hours. Maybe more on a "spare"
      ancient machine (a crime against primality!) that I've not powered up for a

      > If you agree with this, reserve a block or two, I'll do the same, and
      > see just where we can get to.

      The more efficient code should be run. I'll put another 168 GHz minutes
      onto a chunk:

      Block 1 : (0 * 2^40) .. (1 * 2^40) - completed (18 digits found) [AM/PC]
      Block 2 : (1 * 2^40) .. (2 * 2^40) - reserved PC
      Block 3 : (2 * 2^40) .. (3 * 2^40)
      Block 4 : (3 * 2^40) .. (4 * 2^40)
      Block 5 : (4 * 2^40) .. (5 * 2^40)

      If you tell me my code's faster, I'll grab a few more. If you tell me yours is
      faster, I'll only look at my new routine, and put my machine back onto LG.

      > BTW, we will need to have a small overlap just in case a sequence is on
      > a block boundary. It might be an idea to start, say, 100 primes before
      > the block and end 100 primes after.

      Yup, that makes sense. Or at least 20 primes does.
      I'll do 0x<N>fffffff000 to 0x<N+2>0000000fff .


      () ASCII ribbon campaign () Hopeless ribbon campaign
      /\ against HTML mail /\ against gratuitous bloodshed

      [stolen with permission from Daniel B. Cristofani]

      Do You Yahoo!?
      Tired of spam? Yahoo! Mail has the best spam protection around
    • Show all 13 messages in this topic