Issue 52328 - Performance: configmanager
Summary: Performance: configmanager
Status: CLOSED FIXED
Alias: None
Product: General
Classification: Code
Component: code (show other issues)
Version: 680m117
Hardware: All All
: P3 Trivial with 1 vote (vote)
Target Milestone: AOO Later
Assignee: Stephan Bergmann
QA Contact: issues@framework
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2005-07-22 13:36 UTC by malte_timmermann
Modified: 2011-01-21 17:40 UTC (History)
1 user (show)

See Also:
Issue Type: DEFECT
Latest Confirmation in: ---
Developer Difficulty: ---


Attachments

Note You need to log in before you can comment on or make changes to this issue.
Description malte_timmermann 2005-07-22 13:36:21 UTC
Some possible space for improvements:

- reduceExpandedForLocale: modify localized props in place instead of copying 
entire structure

- LocalStratumBase::isMoreRecent() optimize for empty query timestamp

- TreeImpl::getDataAccessor - reuse same accessor, if we are in the same 
segment

- CacheController::addComponentData - translation from INode-data to flat data is 
expensive
Comment 1 joerg.barfurth 2005-10-10 10:11:06 UTC
Handing over

@sb: In this issue I noted some more or less low-hanging fruit for optimizing
performance, which I found by looking at MT's profiling data. It is more a
tracking issue. An individual issue should be filed for each individual problem
as it is addressed.

Of the listed items, the last one is a rather large change and a serious amount
of work. The isMoreRecent change can avoid a lot of unnecessary file stats (but
these files are stated later during the cache check, so the gain is not as big
as one could hope). The other changes eliminate unneeded (probably) allocations
(reduce...Locale) and mutexes (the allocator one).

For more details get back to me.
Comment 2 Stephan Bergmann 2005-10-11 08:02:17 UTC
.
Comment 3 mmeeks 2007-01-17 11:26:38 UTC
The 'Accessor' issue is nailed in CWS configrefactor01 - since we remove
'Accessor' altogether :-) It's interesting though, that my profiles (admittedly
on the re-factored code), don't show any of these areas as significant - but
perhaps I'm missing something (?)

Also - Mutex wise, I've CWS configrefactor01 removes ~all locking in favour of
one big lock, which substantially simplifies the code (& accelerates it).
Comment 4 utomo99 2007-06-24 05:36:58 UTC
.
Comment 5 Stephan Bergmann 2011-01-21 17:33:56 UTC
meanwhile, configmgr code has been re-written completely
Comment 6 Stephan Bergmann 2011-01-21 17:40:03 UTC
.