RESOLVED FIXED 39337
Move Attr pointer from Attribute to an out of band HashMap 
https://bugs.webkit.org/show_bug.cgi?id=39337
Summary Move Attr pointer from Attribute to an out of band HashMap
Sam Weinig
Reported 2010-05-18 18:28:22 PDT
Attrs are rare enough that it does not make sense to keep them in the Attribute itself. We should move them to a global rare data HashMap.
Attachments
patch (5.18 KB, patch)
2010-05-18 18:31 PDT, Sam Weinig 		no flags
DetailsFormatted DiffDiff
patch 2 (5.29 KB, patch)
2010-05-18 18:47 PDT, Sam Weinig 		mjs: review+
DetailsFormatted DiffDiff
Show Obsolete (1) View All Add attachment proposed patch, testcase, etc.
Sam Weinig
Comment 1 2010-05-18 18:29:50 PDT
<rdar://problem/8001168>
Sam Weinig
Comment 2 2010-05-18 18:31:40 PDT
Created attachment 56437 [details] patch
Early Warning System Bot
Comment 3 2010-05-18 18:43:52 PDT
Attachment 56437 [details] did not build on qt: Build output: http://webkit-commit-queue.appspot.com/results/2321307
Sam Weinig
Comment 4 2010-05-18 18:46:28 PDT
This patch doesn't build it seems due to a stupid merging issue. The fix is removing initialization of m_impl(0). Will make a new patch.
Sam Weinig
Comment 5 2010-05-18 18:47:05 PDT
Created attachment 56438 [details] patch 2
Maciej Stachowiak
Comment 6 2010-05-18 18:52:54 PDT
Comment on attachment 56438 [details] patch 2 r=me
Sam Weinig
Comment 7 2010-05-18 18:57:08 PDT
Landed in r59731.
Alexey Proskuryakov
Comment 8 2010-05-19 11:00:21 PDT
What effect does it have on XPath performance? One test to try would be <http://ejohn.org/apps/classname/xpath.html>.
Sam Weinig
Comment 9 2010-05-19 17:25:44 PDT
(In reply to comment #8) > What effect does it have on XPath performance? One test to try would be <http://ejohn.org/apps/classname/xpath.html>. It is around 0.2-0.3% loss, which should be acceptable given the big win on memory.
Note You need to log in before you can comment on or make changes to this bug.