On Sun, Jun 18, 2006, Ross Walker wrote:
>
> Now PME scales as N ln N so we would expect a 100 million atom simulation to
> take approximately (N=100*10^6/408000) = 245.1 * ln 245.1 = 1350 times more
> computation.
Actual scaling, especially for a 12 Ang. vdw cutoff(!) is probably closer to N
than to NlnN. If so, one would gain about a factor of 5 in your calculation,
getting down to more like 1 Pflop.
...dac
Received on Wed Jun 21 2006 - 06:07:09 PDT
