[quote author="mbrasser" date="1293060738"]In general, QML will always try to draw things the fastest way. Often this means that we internally cache things to a pixmap (for example, for 4.7.0 all Text is internally cached to a pixmap, though in 4.7.2 we will switch to using a QStaticText-like solution, which retains good performance and uses less memory). In terms of drawing a vertical linear gradient in a plain rectangle, that can be optimized quite well without caching, so we don't cache to a pixmap. (A gradient in a rounded rect, or bordered rect, however, is not well optimized)
There are certainly cases when using your own pre-composed image is the best way to go, see for example http://doc.qt.nokia.com/4.7-snapshot/qdeclarativeperformance.html#image-resources-over-composition.
[/quote]
Thanks for the explanation! Interesting that it seems to contradict the advice given at the Dev Days on this, but perhaps I just misunderstood. Still, I am very glad that there is constant work being done to improve performance. I guess in the end, there is no substitute for just measuring what performs better for your own specific case.