[webkit-qt] Performance regression in the image_cycling

Holger Freyther zecke at selfish.org
Sun Feb 28 01:27:53 PST 2010


On Friday 26 February 2010 19:47:57 Holger Freyther wrote:
> Hi All,
> 
> I was using some of my spare time to compile Qt origin/master and I was
> executing my image_cycling test case.
> 
> I have compared a Qt4.6 build with my public image branch, with a plain 4.7
> build. In both cases same Qt configure switches were used.
> 
> One iteration on the test went from 38msec to 104msec. It would be nice if
> someone could confirm this measurement.

My initial idea would be it is the QGifHandler::imageCount patch. From a quick 
look the new scan method will remember the sizes of the images and can 
determine the imageCount... but if I ask for image #6 it probably starts from 
the beginning and decodes everything until it...

When scanning the image it should remember the offset in the file where the 
image is starting and it should jump there.... The WebCore GIF decoder is 
doing it that way.

it would be really nice if you could escalate it within Nokia.


z.




More information about the webkit-qt mailing list