RESOLVED FIXED Bug 38820
Further optimization for SVG's repaintRect calculation 
https://bugs.webkit.org/show_bug.cgi?id=38820
Summary Further optimization for SVG's repaintRect calculation
Dirk Schulze
Reported 2010-05-09 13:26:59 PDT
Masking: We don't have intelligent algorithms, to determin the smallest size of a masked objekt to get a smaller repaintRect in SVG. ClipPath: We calculate the union of all childs of a clipPath every time, the clipPath is used by a target. This union can be cached and transformed by the targets objectBoundingbox on calling resourceBoundingBox(). This is no lost if only one target takes the clipPath resource, but will speed up the calculation if more than one object use the same resource. This also avoids multiple transformations by localToParentTransform and the follwoing mapRect call. The same algorithm can be used by masking too.
Attachments
Patch (50.02 KB, patch)
2010-05-09 13:47 PDT, Dirk Schulze 		no flags
DetailsFormatted DiffDiff
View All Add attachment proposed patch, testcase, etc.
Dirk Schulze
Comment 1 2010-05-09 13:47:52 PDT
Created attachment 55508 [details] Patch
Eric Seidel (no email)
Comment 2 2010-05-09 13:52:01 PDT
Comment on attachment 55508 [details] Patch We're certain this changed no pixel tests, correct?
Dirk Schulze
Comment 3 2010-05-09 14:04:26 PDT
(In reply to comment #2) > (From update of attachment 55508 [details]) > We're certain this changed no pixel tests, correct? I allways test with pixel tests and a tolerance of 0. No pixel changes.
Nikolas Zimmermann
Comment 4 2010-05-10 01:51:38 PDT
Comment on attachment 55508 [details] Patch Dirk++, r=me.
Dirk Schulze
Comment 5 2010-05-10 08:59:27 PDT
Comment on attachment 55508 [details] Patch Clearing flags on attachment: 55508 Committed r59081: <http://trac.webkit.org/changeset/59081>
Dirk Schulze
Comment 6 2010-05-10 08:59:37 PDT
All reviewed patches have been landed. Closing bug.
Note You need to log in before you can comment on or make changes to this bug.