--- On Fri, 3/15/13, djbroadhurst <d.broadhurst@...
> Phil Carmody <thefatphil@...> wrote:
> > I'm pretty sure that MJOS & I announced out record-breaking
> > Carmichaels either here (or on NMBRTHRY?) whenever it was
> > (8-10 years ago?).
Wow, if it was 200 times larger, then we did push it all the way to its limits. I'd forgotten that Markku-Juhani had done that, I don't have his logs. He did have access to part of what was once the 7th largest machine in the world when he was found that. A bit of a rewrite (we'd need more than 32 bits) would have been necessary to go any further. 6 Moore's Law doublings does indicate that a 6-billion factor Carmichael should be possible with the same algorithm on a cluster in 2013. With more-distributed computing, perhaps even more. The long boring bit that grows at O(n) is embarassingly parallel, and the "clever" bit (what Loh overlooked) only grows at O(sqrt(n)).