[Webkit-unassigned] [Bug 29279] [Qt] Use RGB16 format for images on Symbian platform

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Wed Oct 14 20:48:00 PDT 2009


--- Comment #28 from Holger Freyther <zecke at selfish.org>  2009-10-14 20:47:59 PDT ---
(In reply to comment #27)

> Can you provide some details about your test case:
> 1. what's the size of your image(s)?
> 2. Without the image, what's the memory usage?
> But anyway, we can calculate the memory saving = w*h*4/2. For example, a
> 800x600 image saves 0.96MB.

I'm using the real web with row01 to row39 and I captured images that got
decoded during rendering on a 1024x768 viewport. I have measured the memory
usage on the reductions/image_cycling in the same tree.

  - http://gitorious.org/qtwebkit/performance/blobs/master/common/common_init.h

> > No. We are fixing things where they need to be fixed. I have no idea about
> > constraints but you will probably have more than one patch to Qt when doing a
> > release/deployment. Just add the API after Qt has agreed that it will be
> > accepted.
> I totally agree adding API as early as possible is great idea. This doesn't
> prevent us doing the implementation at webkit level for now. Otherwise, we have
> to wait for mid next year to get the job done.

One thing after another. Currently this will only add a negative performance 
impact on converting to 16bit in Qt/ImageDecoder and converting back to 32bit
in QPixmap::fromImage.

So let us focus on:

1.) Showing that the color conversion from RGB32 to RGB16 is actually saving
significant amount of memory and then to look at the performance impact of that

2.) Look at decoding gif's to 8bit or below...

> > Currently I'm not really convinced that the 1.7MB is a big difference, do you
> > have other data showing a bigger saving?
> I won't think the memory saving will the different as long as we find out three
> figures:
> the "baseline": memory consumption with no image; memory consumption with image
> using this 16bit conversion and without this conversion.

For the baseline. Looking at the mirrored sites, do you think anything major is
missing? Should we have a pure top50 data set as well. I highly encourage you
start to look into the mirror tool and attempt to verify my benchmark.

Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

More information about the webkit-unassigned mailing list