I have been having fun with prime numbers. I sat down and, following a hunch and after a few weeks of headbanging against the wall, I was able to write an algorithm that, at least on my local, provides better results than the Sieve of Atkin (at least, faster than the implementation that I found to test against).
Right now it's written on java but I will probably sit down during the weekend to translate it into C to measure it up against Atkin there as well.
In case you want to give it a test-drive:
https://github.com/eantoranz/patternsieve
Master only holds 2 classes: - the "reference" implementation of the algorithm (which is much slower than Atkin, by the way) - the version that I ended up having after steroids (but which holds the same concept as the reference implementation at its heart)
Then there's the 'compare' branch where I wrote a couple of additional java classes that can be used to easily and readily run tests plus an implementation of Eratosthenes and Sieve of Atkin (credits provided in commits and in code... and in output when using the classes).
What would be the best steps to follow so that people knows about it and give it a round of tests? Just writing about it over here should be enough to get people to notice it? Perhaps I should mail someone in the math community? I'm kind of a rookie in terms of "how to do and publish research".
PS By the way.... what do I mean with "faster"? I think overall it can provide some 30% faster results (but it varies a lot because of the way that the algorithm works so it can be very close to Atkin or provide more than 50% better results... depending on how many primes are being searched).
I just hope I don't get a lot of negative votes and perhaps get some people interested.