On 10/18/2010 11:11 AM, Ingo Molnar wrote:
* H. Peter Anvin <hpa(a)zytor.com> wrote:
> On 10/18/2010 09:48 AM, Eric Paris wrote:
>> 1) IMA uses radix trees which end up wasting 500 bytes per inode because the key
>> is too sparse. I've got a patch which uses an rbtree instead I'm testing
>> will send along shortly. I found it funny working on the patch to see that
>> Documentation/rbtree.txt says "This differs from radix trees (which are used
>> efficiently store sparse arrays and thus use long integer indexes to
>> insert/access/delete nodes)" Which flys in the face of this report.
> Radix trees can efficiently store data associated with sparse keys *as long as the
> keys are clustered*. For random key distributions, they perform horribly.
For random key distributions hash and rbtree data structures are pretty good
But the (much) more fundamental question is to turn the non-trivial allocation
overhead of this opt-in feature into truly opt-in overhead.
Yes, and not just the allocation overhead, but apparently locking