On Nov 23, 2012, at 1:27 PM, Robert Knight <[email protected]> wrote:
>> The main reason for committing this now is to close the feature regression >> from Qt 4 to Qt 5. >> Qt 4 gets high-dpi support via the CoreGraphics paint engine. Qt 5 uses >> raster and we need to do the implementation work in Qt. > > Have you measured at all how performance compares between Qt 4 and Qt > 5 in HiDPI mode? I assume CoreGraphics benefits from hardware > acceleration and although Qt's raster engine has often been faster > than the 'native' engine, I wonder if that will still be the case when > processing 4x the number of pixels on a Retina-era Mac. That's a good question. I've tested a bit with Qt Creator and the Qt 4 version certainly feels more performant at this point. I'm not 100% happy with he current backing store implementation on Mac so I hope this is something we can improve. (There are already changes in the pipeline, see https://codereview.qt-project.org/#change,40445) Conceptually I don't see a reason why raster should be slower than CoreGraphcs since both has to do the same job: draw using CPU and upload to the graphics card. (CoreGraphics has the QuartzGL mode where drawing is done using shaders on the GPU, but this is disabled by default.) Finally, we ran Qtify (the Spotify clone) from the desktop components dev days presentation on a retina Mac with really good performance. This is a Qt Quick (2) application built on OpenGL and scene graph instead of the traditional raster code path. Morten _______________________________________________ Development mailing list [email protected] http://lists.qt-project.org/mailman/listinfo/development
