[Qt5-feedback] QML and software renderer
gunnar.sletta at nokia.com
gunnar.sletta at nokia.com
Fri May 13 07:22:11 CEST 2011
On May 12, 2011, at 12:18 AM, ext Andrey Batyiev wrote:
> Hello
>
> This is crosspost from the blog, because comments are closed.
>
>
> It would be great to have Qt-approved software renderer, so it would
> make all rendering stuff on CPU like before.
This appeared in another thread too, and I think that a software renderer is the way to solve the "problematic GL" issues, yes. That begs the question of where the software renderer comes from. Will llvmpipe get there? Who will pitch in to get it there? I do not have answers to that today.
> Advantages:
> 1. ability to compare between QWidget and QML rendering
> 2. releasing fears like “my platform doesn’t support opengl es 2.0″
> and providing vital solution for unfortunate
> 3. remote sessions will work like before
>
> Disadvantages:
> 1. emulation should be *perfect*, otherwise it would destroy performance of QML
>
>
> Am I right that QWidget stuff is rendered on CPU and should be
> comparable to QML with software renderer ( = rendering on CPU)?
> What do you think?
Windows is rendering all widgets using software and has been doing so since 4.0. On X11 there has been a growing trend to use software for some time, but its not on by default in 4.7. From 4.8, software rendering is on by default on all desktop platforms. That means all Qt platforms render equal out of the box.
Software rendering of GL should not be much different if done the right way. Add to that that the QML scene graph can do full scene optimizations and discard pixels behind opaque objects, I would expect performance to be better, but it remains to be seen.
-
Gunnar
More information about the Qt5-feedback
mailing list