On Thu, 3 Sep 1998, David S. Miller wrote:
> No way, AVL is not going back into the kernel for vmas, ever.
So, we release a 2.2 which for certain tasks is an order of magnitude
slower than 2.0?
For both extremes it has been shown that this is not the case for a
fuzzy hash implementation, _AND_ (the most important part, and the
reason why AVL was yanked in the first place by Linus and myself)
fuzzy hash has the property that it has lower latency for the case
that matters %99 of the time which is only a few mappings and nothing
sardonic like Electric Fence.
The fuzzy hash implementation has only been slightly tuned by Audrey,
I am sure it can be made to run with even lower latencies than his
second patch set does.
The reason Linus and I yanked it in the first place was to not go back
to it. The tree balancing cost for the common case is just gross.
Look at the test results posted here before making such statements
about all non-AVL methods.
Later,
David S. Miller
davem@dm.cobaltmicro.com
-
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/faq.html