Thanks Reaud and Mathew! The s/update/updateGL/ fixed the problem! Btw, I initially copied the use of updateGL() from .../examples/opengl/hellogl/glwidget.cpp, which still (git HEAD) uses it. Shouldn't this be changed for the same reason as well?
Regards, Dov On Fri, Nov 21, 2014 at 11:39 PM, Matthew Woehlke < mw_tr...@users.sourceforge.net> wrote: > On 2014-11-21 05:22, Renaud wrote: > > I saw in your small example that you use "updateGL()", I suggest you to > > change it to "update()". > > > > It may help the performance on Qt5. > > Hoo boy... I'm not sure why this would work better on Qt4 (see possible > idea, below), but updateGL() *IMMEDIATELY* calls your paintGL, i.e. > forces an immediate render to occur and don't return until the render is > completed. You *almost never* want to do that (if you're about to grab > the buffer is probably the only time you would). If you do that in a > mouse event handler, then (referring to your original observation) you > will indeed get a render for every mouse event that is processed. > > update() on the other hand tells Qt that your widget needs to be > repainted "some time". In particular, "some time" generally means "when > you're done processing outstanding events". This will let you handle a > bunch of mouse events, do whatever changes you need, and then paint > (render) once all of them have been handled. > > It may be that in Qt4 there is some coalescing of mouse events happening > that is not happening in Qt5. > > At any rate, you will probably get much more similar performance with > update() vs. updateGL(). > > -- > Matthew > > _______________________________________________ > Development mailing list > Development@qt-project.org > http://lists.qt-project.org/mailman/listinfo/development >
_______________________________________________ Development mailing list Development@qt-project.org http://lists.qt-project.org/mailman/listinfo/development
