Review Request 119524: Force OpenGLES if on Qt 5.4

[Sebastian Kügler]

> On July 29, 2014, 2:01 p.m., Sebastian Kügler wrote:
> > I'm all for this, as it means we reduce the set of features we use from the underlying stack. We're even focusing on the features that will get more testing (since it's mobile, and that's the hype). I consider it an advantage if we can use a more restricted API, and communicate that we don't care about "random OpenGL extensions", instead asking "make sure the OpenGL-ES implementation" is sound and works for us.
> > 
> > It would be useful to hear from Martin if there's any technical problem with this strategy. We should also align this kind of decisions with KWin. (Perhaps the same is a serious option for KWin?)
> > 
> > From my personal experience, I've tested the OpenGL-ES options we had for some time, and never ran into a regression caused by moving from OpenGL to OpenGL-ES.
> 
> David Edmundson wrote:
>     Martin G shares his thoughts here: https://bugreports.qt-project.org/browse/QTBUG-40341
>     I'm not sure I understand all of it though :)

Well, he doesn't answer the question of this RR. :)

IMO, we can do two things:
- merge, cross fingers, watch out for regressions, when Martin gets to it, ask him if it's a sensible thing
- wait for Martin, then decide to merge or not

I'm cool with with either strategy. (Would be quite interesting though to see what happens when we switch. :))


- Sebastian


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://git.reviewboard.kde.org/r/119524/#review63437
-----------------------------------------------------------


On July 28, 2014, 6:09 p.m., Aleix Pol Gonzalez wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://git.reviewboard.kde.org/r/119524/
> -----------------------------------------------------------
> 
> (Updated July 28, 2014, 6:09 p.m.)
> 
> 
> Review request for Plasma.
> 
> 
> Repository: plasma-workspace
> 
> 
> Description
> -------
> 
> When trying to reduce memory usage in plasma shell, we realized that one of the reasons we got such big memory footprint was actually the intel driver (on intel systems, that is). After some investigation, David found out we were going through some memory-consuming path [1]. A way to workaround it is by using only OpenGLES. Everything still works here after this patch, so it seems to be a good change to get in.
> 
> This change needs Qt 5.4, to get the new API that lets us enforce a QSurfaceFormat, hence having it ifdef'd.
> I'm quite unaware of problems we might find. Knowing we probably want to work in different embedded devices, suggests that OpenGLES on all platforms sounds like a safe bet.
> 
> [1] http://comments.gmane.org/gmane.comp.video.mesa3d.devel/60848
> 
> 
> Diffs
> -----
> 
>   shell/main.cpp e34578d 
> 
> Diff: https://git.reviewboard.kde.org/r/119524/diff/
> 
> 
> Testing
> -------
> 
> I'm using it now, without visible problems. Now we see no trace of i915 in the massif reports.
> 
> 
> File Attachments
> ----------------
> 
> before
>   https://git.reviewboard.kde.org/media/uploaded/files/2014/07/28/5a6beb4f-24c0-4cee-a7a2-038385e35119__plasma-massif-opengl.png
> after
>   https://git.reviewboard.kde.org/media/uploaded/files/2014/07/28/fc301ff5-cb1d-4654-a57f-82990220c8e3__plasma-massif-opengles.png
> 
> 
> Thanks,
> 
> Aleix Pol Gonzalez
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/plasma-devel/attachments/20140729/03e1aa16/attachment.html>