Re: Prime counting algorithms, speculation

From: Christian Bau (christian.bau_at_cbau.freeserve.co.uk)
Date: 08/04/04 

Date: Thu, 05 Aug 2004 00:47:49 +0100

In article <3c65f87.0408041019.62cf943d@posting.google.com>,
 jstevh@msn.com (James Harris) wrote:

> jurgenr@web.de (Jürgen R.) wrote in message
> news:<4110edb4.28774195@news.individual.de>...
> > [...]
> > >
> > >My fastest prime counting program is PrimeCountH.java, which readers
> > >can easily pull up by a Google search or a Google Groups search, and
> > >it blows away ALL the times that you'll find posted on that page.
> > >
> > [... on and on]
> >
> > Just tell us how fast it is: To count the primes < N there is some
> > function F(N) such that the time taken by your program is less than
> > C*F(N) for some fixed C. Tell us what F is. If you don't know stop
> > taking about how fast your algorithm is, because in that case it is of
> > no interest at all.
> >
> > Jürgen
>
> Some sci.math poster put up a webpage claiming that I entered a
> competition to build fast prime counting programs and lost badly, when
> in fact, I never entered any competition, and actually my fastest
> implementation--PrimeCountH.java--is not only far faster than anything
> he presented on his webpage, it's been available on-line for quite
> some time.

You claimed that your prime counting program blows away everything else.
To me, that sounds like entering a competition. Maybe you didn't
actually enter a competition, but the source code that can be found at
www.cbau.freeserve.co.uk runs about 1000 times faster for large N than
the fastest algorithm you ever published, and it gives correct results
for values of N up to about 1.6 * 10^19, where your algorithm just runs
out of memory. So if you ever decide to enter a competition you have
some serious work to do.

That's the nice thing about software: People can check whether it works
or not, and they can check how fast it runs, and your usual bullshitting
doesn't help you at all.

Quantcast