B+tree discussions

[Robert Collins]

On Thu, 2008-07-03 at 17:09 +1000, Robert Collins wrote:
> iter_random_one was spectularly slow. 60% of it was in split-lines - so
> here with figures from using a C parser.
> C parser:

I've also upped the leaf cache node size to 4MB rather than 400K
(currently per index). As already noted this really impacts the tests
that were cache thrashing (but I prefer cache thrashing to VM thrashing
- cache thrashing is faster than hitting every page in swap to do
refcount checks/adjustments).

In short - faster or equal across the board:

BTreeIndex: iter_all_entries in 1.026
GraphIndex: iter_all_entries in 10.948

BTreeIndex: iter_random_one in 12.712,
GraphIndex: iter_random_one in 12.052,

BTreeIndex: get_components_positions(2000) in 7.057,
GraphIndex: get_components_positions(2000) in 22.545,

BTreeIndex: search_from_revid in 3.499 found 18130,
GraphIndex: search_from_revid in 7.201 found 18130,

BTreeIndex: miss torture in 22.602,
GraphIndex: miss torture in 365.309,

-Rob
-- 
GPG key available at: <http://www.robertcollins.net/keys.txt>.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20080703/5222d689/attachment.pgp