[webkit-qt] Performance regression

[Coda Highland]

Great, thanks for the pointers. That did solve the problem. Now to figure
out why the opacity is broken...

/s/ Adam

On Thu, Mar 22, 2012 at 11:21 AM, <viatcheslav.ostapenko at nokia.com> wrote:

>  Make sure that accelerated compositing and tiled backing store are off. *
> ***
>
> QWebSettings::AcceleratedCompositingEnabled****
>
> QWebSettings::TiledBackingStoreEnabled****
>
> ** **
>
> There was a lot of development in this area and I’m not sure what was
> decided about defaults. For QWebView this should be off for sure.****
>
> Also, try to increase object cache sizes:****
>
> *QWebSettings::setObjectCacheCapacities<http://qt-project.org/doc/qt-4.8/qwebsettings.html#setObjectCacheCapacities>
> * ( int *cacheMinDeadCapacity*, int *cacheMaxDead*, int *totalCapacity* )*
> ***
>
> ** **
>
> Sl****
>
>
> On 03/22/2012 11:04 AM, ext Coda Highland wrote: ****
>
> Just ordinary QWebView. The only variations I've tried are to switch
> between -graphicssystem raster and -graphicssystem native, and the
> performance is roughly the same in both. (Raster has better visual
> rendering though, but that seems to be a different regression in 4.8.0.) *
> ***
>
> ** **
>
> /s/ Adam****
>
> On Thu, Mar 22, 2012 at 9:19 AM, <viatcheslav.ostapenko at nokia.com> wrote:*
> ***
>
> Could you give more details how do you use webkit.****
>
> Is it QWebView or QGraphicsWebView?****
>
> If QGraphicsWebView:****
>
> Do you have graphics item cache on or off?****
>
> Do you have tiled backing store on or off?****
>
> Do you have accelerated compositing on or off?****
>
> Di you use GL widget for GraphicsView or not?****
>
>  ****
>
> Thanks****
>
>  ****
>
>  ****
>
> *From:* webkit-qt-bounces at lists.webkit.org [mailto:
> webkit-qt-bounces at lists.webkit.org] *On Behalf Of *ext Coda Highland
> *Sent:* Thursday, March 22, 2012 12:47 AM
> *To:* webkit-qt at lists.webkit.org
> *Subject:* [webkit-qt] Performance regression****
>
>  ****
>
> I'm on Mac (Snow Leopard), and building 32-bit (Carbon, of course) because
> of other dependencies.****
>
>  ****
>
> Back in Qt 4.7 (specifically, on August 14th, 2011), I had built the
> QtWebKit 2.2 branch from trunk in order to get some features I needed for
> my application.****
>
>  ****
>
> Today I got tired of dealing with the broken inspector and decided to
> update to 4.8.0. Unfortunately, the version of QtWebKit bundled with 4.8.0,
> as well as a fresh checkout of trunk, have severe performance regressions,
> especially with scrolling. I kept Qt 4.8.0 but switched back to my old
> build of QtWebKit from August 14th and the performance returned to normal.
> (I still had another regression regarding opacity, but that seems to be a
> Carbon-specific regression in Qt itself, not QtWebKit.)****
>
>  ****
>
> I'm not looking forward to spending my weekend bisecting this; is this a
> known issue? Does anyone have any pointers on what change(s) might be
> relevant?****
>
>  ****
>
> Thanks,****
>
> /s/ Adam****
>
> ** **
>
> ** **
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-qt/attachments/20120322/4b3b9006/attachment.html>