Daniel Russel wrote: > O(N) is not is irrelevant. The constants are what matters and they > were bad. The code I submitted is better, but still not as fast as the > CGAL one.
I must be misunderstanding your benchmark, in that case. You say: Grid For 50 particles got 3.63301706314 For 100 particles got 3.84657287598 For 1000 particles got 6.80274581909 For 10000 particles got 11.4696931839
and that the results are normalized by the number of particles. Surely if it were O(N) you'd get about 3.63 regardless of the number of particles, since by 'normalized' you mean these numbers are already divided by N?
>> All patches to IMP need to be reviewed before they are checked in. > Not that we could do anything even if we did review them...
If Keren, for example, were to review your patch and say it looks OK, I could check it in very quickly even over a dodgy dialup connection. Or others can have commit rights. It just doesn't make sense for the same person writing the code to commit it.
> Aren't you back now?
Monday. Today is a weekend.
> I thought LGPL could be linked against anything. And I thought IMP was > going LGPL for exactly this sort of reason. It was a while ago, so I > don't remember the final decision.
Nope, the other way around: anything can be linked to LGPL. We still have to respect the licenses of our dependencies. LGPL is what I proposed for IMP, but Andrej has not yet made a decision on this score. I proposed LGPL so that others could write their own extension modules and not be forced to make them open source.
Ben