Re: [PrimeNumbers] Re: Record-setters for prime pair-concatenations of 4 cons. primes
- [bringing back on-list]
--- On Mon, 2/7/11, James Merickel <merk7777777@...> wrote:
> Hi, Phil. What you have there
> as the first 10/12 agrees with my curio at Prime Curios.
Excellent. Double-checking is always good for both parties (and third parties who have to trust the results).
> I thought the nice first 9/12 I described
> would have been accepted, but only this 10/12 was.
> Your search is clearly around a full order of magnitude
> faster than mine, and if you restricted it to searching for
> 11/12 it would be faster still.
With the current algorithm, there's no direct increase in restricting it to 11/12. I don't do any trial-division, I just test each concatenated number in turn (with GMP, so that it does some trial division). Most of the time it doesn't even get to the 10th number, so stopping once 10 is possible, but 11 is impossible, almost never happens. So no optimisation possible. That's what you get with rushed code using nothing but brute force.
However, that's because my algorithm's a bit dumb. I should add a pre-sieving stage. If I do that, then firstly it should be a fair bit faster, and secondly capable of the optimisation you mention.
> Unfortunately, I don't
> think I have the programming software. I could perhaps
> download it to disk at the library in order to install it on
> my computer, but I'm a little tied up with other things
> right now.
An executable made by me won't help you. I use only linux, and my development machine is a PowerPC. My other development machine is ARM-based. I don't do PeeCees and Windoze at all. However, if I publish the code, others can certainly build it for you. When I last had a windows machine, I seem to remember "cygwin" being pretty handy, perhaps you could install that.