NEW 116599
Performance regression in HitTest 
https://bugs.webkit.org/show_bug.cgi?id=116599
Summary Performance regression in HitTest
Xan Lopez
Reported 2013-05-22 02:31:47 PDT
(This title should become more specific at some point) The patch in bug #85849 seems to have caused a regression in the performance of hit testing under some circumstances. I'm not very familiar with this code, but looking at the patch it seems to change a O(N) loop without recursive hit tests for two separated loops with recursive hit tests, one of them being O(N*M) (N being the siblings of a given node, M its parents). I attach a layout test that timeouts with trunk, but that does not with the patch in that bug reverted.
Attachments
hittest.html (493.83 KB, text/html)
2013-05-22 02:34 PDT, Xan Lopez 		no flags
Details
View All Add attachment proposed patch, testcase, etc.
Xan Lopez
Comment 1 2013-05-22 02:34:12 PDT
Created attachment 202511 [details] hittest.html This layout test time outs for me. It does just one hit test, so it really should not. It's a reduction of a real world page (Astronomy picture of the day), which is where we found this regression.
Note You need to log in before you can comment on or make changes to this bug.