Page 1 of 2 12 LastLast
Results 1 to 10 of 32

Thread: 2D Performance Also Impacted By Unity On XMir

Hybrid View

  1. #1
    Join Date
    Jan 2007
    Posts
    14,350

    Default 2D Performance Also Impacted By Unity On XMir

    Phoronix: 2D Performance Also Impacted By Unity On XMir

    Earlier today I delivered the first benchmarks of Ubuntu's Unity 7 running over XMir to run the current X11 desktop atop the Mir Display Server via this compatibility layer. These benchmarks documented the performance impact of running OpenGL games when having to deal with XMir rather than just a clean X.Org Server running on the hardware. The extra step in the rendering process did result in a measurable performance impact, especially when the performance of the open-source Linux graphics drivers is already lower than their proprietary brethren. The benchmarks to now show illustrate that the 2D rendering performance also takes a hit when running Unity on XMir.

    http://www.phoronix.com/vr.php?view=18832

  2. #2
    Join Date
    May 2012
    Posts
    762

    Default

    But what's the 2D performance on pure Mir? Faster, same, worse than on X.org?

    After all, we're all looking forward to use pure Mir or Wayland as much as possible, with X.org on top of them as little as possible.

  3. #3
    Join Date
    Jan 2011
    Posts
    1,287

    Default

    Quote Originally Posted by mark45 View Post
    But what's the 2D performance on pure Mir? Faster, same, worse than on X.org?

    After all, we're all looking forward to use pure Mir or Wayland as much as possible, with X.org on top of them as little as possible.
    Yes, but currently there is nothing where it's easy to perform benchmarks of native Mir (i.e., at least an alpha distro using it directly).
    Also, that's what will be used by default in both 13.10 and 14.04, so that comparison is much more interesting in the short term than the native one, from a user's perspective.

  4. #4
    Join Date
    Mar 2011
    Posts
    209

    Default What about X vs XMir vs XWayland benchmarks?

    Those results should be interesting.

  5. #5
    Join Date
    Apr 2011
    Posts
    153

    Default

    Quote Originally Posted by M1kkko View Post
    Those results should be interesting.
    I don't think there will be much difference.
    At least wayland will be natively supported by qt4,5/kde and gtk!

  6. #6
    Join Date
    Jul 2010
    Posts
    591

    Default

    Quote Originally Posted by kokoko3k View Post
    At least wayland will be natively supported by qt4,5/kde and gtk!
    Wayland will be supported natively by Qt 5.2 and later; not by Qt 4.

  7. #7

    Default

    Quote Originally Posted by mark45 View Post
    But what's the 2D performance on pure Mir? Faster, same, worse than on X.org?

    After all, we're all looking forward to use pure Mir or Wayland as much as possible, with X.org on top of them as little as possible.
    Even worse. If we use cairo as a yard stick for measuring acceleration across the different solutions, then cairo-xlib is many times faster (when supported by a good driver i.e. sna or nvidia, and even when supported by bad drivers such as uxa and fglrx) than cairo-gl (with their respective OpenGL drivers). We are back to looking at using the CPU and pushing images around.

    Pure client side rendering also has higher memory overheads as what the display server does cache between multiple clients, is now allocated separately in every client.
    Last edited by ickle; 06-29-2013 at 03:04 AM.

  8. #8
    Join Date
    Apr 2011
    Posts
    153

    Default

    Quote Originally Posted by ickle View Post
    Even worse. If we use cairo as a yard stick for measuring acceleration across the different solutions, then cairo-xlib is many times faster (when supported by a good driver i.e. sna or nvidia
    (OT)
    Unfortunately proprietary nvidia drivers performs really poor when it comes to draw gradients through cairo.
    ...At the point that if you scroll a webpage containing native widgets (firefox) and your theme uses gradients for them, you have a very choppy scrolling.
    Cairo needs then to be patched to explicitely rely on CPU only for such operations.
    Works good with OSS drivers instead.

  9. #9
    Join Date
    Jul 2009
    Posts
    241

    Default

    anyone with G945 intel driver should be aware that due to the newly gained features of mesa 9.2 (opengl 2.1) it takes about half a minute to open unity dash (pushing the super key) ... i guess..
    and its quite interesting that on the intel driver the SW mouse popinter does not move whereas it does move on nouveau nv50 and radeon r500

  10. #10
    Join Date
    Mar 2013
    Posts
    160

    Default

    Quote Originally Posted by BO$$ View Post
    OK so probably in 13.10 we'll take a performance hit. Hopefully the impact will be smaller, much smaller than it is today.... Get to work Canonical, only a few months left. Get it right a lot of people will think better of you. Get it wrong and haters will complain about this performance loss for another decade, even if it will have been fixed in 14.04.
    http://www.olli-ries.com/first-mir-benchmarks/

    One of the reasons for this result set is missing composite bypassing support, which we are aware of since January. Composite bypass helps when apps/benchmarks run fullscreen because… well, because they don’t need to be composited. Gamers out there… there is hope and a plan in place to get you your precious FPS back. This feature/bug is currently scheduled once other key functionality landed. Also, in order to make FPS based benchmarks really count, we need eglSwapInterval(0) implemented, which is currently in progress

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •