Re: AVL trees vs. Red-Black trees

Jamie Lokier (lkd@tantalophile.demon.co.uk)
Sun, 28 Nov 1999 17:44:22 +0100

Oliver Xymoron wrote:
> > The self-adjusting property
> > should theoretically provide cache-like behaviour when most of the
> > lookups are to relatively few nodes in the tree. The downside is that
> > the tree is modified by lookups.
>
> Unfortunately, that means that there's no way it can even begin to scale
> on SMP - all "readers" need exclusive locks unless you can somehow
> guarantee that there'll only ever be one reader/writer. In some instances,
> it might make sense to have completely unlocked per-processor trees
> mapping onto shared structures, but it seems a little painful.

I didn't investigate, but I have a feeling it's possible to do shared
readers on these trees if the pointer changes are done in the right
order.

-- Jamie

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/