Andres Kroonmaa wrote:
> Actually one large blob is used only for holding pointers to actual
> heap nodes. Each heap node is alloced individually per storeentry.
> Today this makes not much of difference anymore. At a time I was
> into it, lru node was part of storeEntry, now its also separate.
> So heap overhead is only about 1-2 pointers higher than lru has.
> It looked much worse back then.
Right.
This is being addressed in the "compactsentry" branch by moving the
policy data back "inline", along with most other "static" information
referenced by the store entries. The use of chunked_mempools will reduce
the need of doing this somewhat (no malloc overhead) but still it is
worthwile I think.
-- HenrikReceived on Wed Oct 03 2001 - 04:22:47 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:14:23 MST