David - The FastSieve module works pretty quickly, I must say. I do not need random numbers on a regular basis, so I can't compare it with other modules. However, computing and printing the first 5 million primes is pretty quick, quick enough that I don't notice an appreciable delay between when I hit Enter and when the results start printing. I would also point out that every single function is written in C++ and there are no need for Perl-side wrappers. Very nice!
For anybody else who might want to try it out, I would point out that David uploaded v 0.02 and my CPAN only wanted to install v 0.01. Unfortunately, 0.01 wants Perl 5.14.2, so it didn't install on my machine. If you have an older Perl (I have 5.14.1), you might have better luck with this command: cpan D/DA/DAVIDO/Math-Prime-FastSieve-0.02.tar.gz The only gripe that I have is that the docs have a slightly unconventional POD structure. The Synopsis section could be improved with comment lines explaining what each function does. However, the docs are complete, so it's really only a minor gripe. Also, this module is a proof-of-concept for Inline::CPP, so the docs don't need to be perfect. Great work! David On Mon, Dec 12, 2011 at 6:48 AM, David Mertens <dcmertens.p...@gmail.com>wrote: > Nice work! I'll have to give it a look later today. > > David > On Dec 12, 2011 1:12 AM, "David Oswald" <daosw...@gmail.com> wrote: > >> I just wanted to get a proof of concept ironed out demonstrating the >> concept of a CPAN module using Inline::CPP (as opposed to some >> old-school XS approach). >> >> Math::Prime::FastSieve is on CPAN now -- or as soon as your local >> mirrors pick it up. It implements the Primes Sieve of Eratosthenes >> using a bit vector in C++, and serves it up to Perl via Inline::CPP. >> >> The steps to making a module with Inline::CPP seem to be pretty much >> exactly as illustrated in the Inline PODs here: >> >> >> http://search.cpan.org/~sisyphus/Inline-0.49/Inline.pod#Writing_Modules_with_Inline >> >> Since Inline::CPP is now passing about 93% of smoke tests (I'll work >> on getting closer to 100% once the dust settles for the current round >> of smoke testing), it's not unreasonable to use Inline::CPP in >> real-world modules. >> >> Oh, and why another Primes module? Because this one beats >> Math::Prime::XS in two important ways: >> 1: It works (well) up to 2.14 billion (M::P::XS craps out around 500M). >> 2: It is at minimum 21% faster than M::P::XS at any given 'n'. >> >> ...plus... >> >> 3: It's an evolution of code used in my presentation at LA Perl >> Mongers and Thousand Oaks Perl Mongers on Inline::C. >> >> >> The slides to my Inline::C presentation (with some discussion on >> Inline::CPP as well) are here: >> http://www.box.com/s/svecm9hq1tam27ggdjds >> >> Dave >> >> -- >> >> David Oswald >> daosw...@gmail.com >> > -- Sent via my carrier pigeon.
