On Friday, September 14, 2012 04:25:18 PM ext Zoltan Horvath wrote: > Okay, then it seems reasonable to roll it out, so I'm going to revert it > next week.
Further discussion with Jocelyn concluded that reverting is indeed the least- effort way of fixing this regression. Simon > On Fri, 14 Sep 2012 16:12:00 +0200, <[email protected]> wrote: > > Forget my last comment, this seems to be directly related to the use of > > QImage instead of QPixmaps. > > > > The directfb platform plugin uses QBlitterPaintEngine which handles > > drawPixmap in QDirectFbBlitter::drawPixmap rather than using straight > > CPU raster for drawImage. > > > > ________________________________________ > > From: [email protected] > > [[email protected]] on behalf of Turcotte Jocelyn > > (Nokia-MP/Oslo) > > Sent: Friday, September 14, 2012 3:53 PM > > To: [email protected]; [email protected] > > Subject: Re: [webkit-qt] [Qt] Strong performance degradation with > > DirectFB since r122720 > > > > Ahh it's a canvas test, this might also be related to the deep-copy > > protection that was removed in the patch, when a QPainter is beginning > > on an existing image. > > > > ________________________________________ > > From: [email protected] > > [[email protected]] on behalf of Turcotte Jocelyn > > (Nokia-MP/Oslo) > > Sent: Friday, September 14, 2012 3:35 PM > > To: [email protected]; [email protected] > > Subject: Re: [webkit-qt] [Qt] Strong performance degradation with > > DirectFB since r122720 > > > > Why was it necessary to switch from QPixmap to QImage? > > > > I don't think that this can be solved on the Qt side, it's a design > > advantage that QPixmap has in this kind of situations. > > As far as I know this might be mostly related to reusing a graphics > > buffer that already has been uploaded to graphics memory. > > > > Jocelyn > > > > ________________________________________ > > From: [email protected] > > [[email protected]] on behalf of ext Zoltan Horvath > > [[email protected]] > > Sent: Friday, September 14, 2012 3:17 PM > > To: [email protected] > > Subject: Re: [webkit-qt] [Qt] Strong performance degradation with > > DirectFB since r122720 > > > > Hi there, > > > > Shouldn't we report this on Qt-side instead? > > > > Cheers, > > Zoltan > > > > > > On Fri, 14 Sep 2012 15:02:13 +0200, Allan Sandfeld Jensen > > > > <[email protected]> wrote: > >> On Friday 14 September 2012, Brianceau, Julien wrote: > >>> With r126284 based webkit version, I get 11,3 displayed fps, with a CPU > >>> usage of more than 95% With r126284 based webkit version with changeset > >>> 122720 reverted, I get 25,6 displayed fps, with a CPU usage of about > >>> 90% > >> > >> Yes with a 50% speed regression also on Qt5, I think we need to revert > >> the > >> patch in trunk, otherwise I would have suggested only reverting it in > >> QtWebKit > >> 2.3. But having as few difference as possible is better, so I am > >> actually > >> relieved it was also slower in Qt5. > >> > >> Cheers > >> `Allan > >> _______________________________________________ > >> webkit-qt mailing list > >> [email protected] > >> http://lists.webkit.org/mailman/listinfo/webkit-qt > > > > _______________________________________________ > > webkit-qt mailing list > > [email protected] > > http://lists.webkit.org/mailman/listinfo/webkit-qt > > _______________________________________________ > > webkit-qt mailing list > > [email protected] > > http://lists.webkit.org/mailman/listinfo/webkit-qt > > _______________________________________________ > > webkit-qt mailing list > > [email protected] > > http://lists.webkit.org/mailman/listinfo/webkit-qt > > _______________________________________________ > webkit-qt mailing list > [email protected] > http://lists.webkit.org/mailman/listinfo/webkit-qt _______________________________________________ webkit-qt mailing list [email protected] http://lists.webkit.org/mailman/listinfo/webkit-qt
