On Thursday 19 September 2013 13:15:16 Mag. Dr. Nikolaus Klepp wrote:
This is done on server side and this engine is very slow. If to switch Qt4 to raster engine, then performance goes to normal.
And that's why raster engine is the default since 4.8 and xrender is no longer present in 5.x. None of these improvements is present in Qt 3 btw, as Qt 3 only supports the slow XRender implementation.
Then what's the explanation for QT4 being way slower than QT3 on e.g. ('cause I have access to this hardware) X61 and GeForce GTX 470?
well first we have to define what "slow" means. Then we have to setup a benchmark which verifies that it is in fact slower. At the moment I don't have any possible way to validate your claim - I have to believe it (no offense, but I do not, see below).
If we can verify that it is in fact slower, then we can look into why that is the case. My personal assumption is that something is incorrectly configured and the newer setup just exposes the problem. Also there's the possibility of bugs in the application you tried or that you compared incomparable things (e.g. Kicker vs Plasma - doesn't make sense, doesn't say anything about Qt).
So let's start with: how do you test that Qt 4 is slower than Qt 3? What's your test, where can I get the source to verify it?
Just as a small reminder: do you really think that can be the case? Qt 4 got trimmed to run on smartphones - I have a few of those (e.g. the N9 and the N950, also used to have a Symbian device) and they would love to have the specs of your X61. Given that Nokia invested lots of money in that area, do you really think the engineers would not have realized that Qt 3 was faster, look at that code and make better use of it in Qt 4?
Cheers Martin