2009/12/21 Greg Stark <gsstark(at)mit(dot)edu>:
> Incidentally there are O(n) algorithms for finding the median (or any
> specific offset). It shouldn't be necessary to sort at all.
it is interesting information. It could to help with missing
optimalisations now.
Pavel
>
> I'm not sure which path this argues for - perhaps Tom's position that
> we need more optimiser infrastructure so we can see how to accomplish
> this. Perhaps it means you should really implement median() with an
> internal selection alagorithm and not depend on the optimizer at all.
>
>
> --
> greg
>