Intel vs. Modesetting X.Org DDX Performance Impact

[Mystro256]

Ok I've got a beginner question here.
I thought that intel's graphics drivers/modules were actually included in the kernel source. So when I install something like xserver-xorg-video-intel, what am I actually installing? are these closed source blobs or something? I didn't think that intel graphics needed anything to be installed. Same with nouveau? What am I missing?
Thanks guys!

It's the xorg usermode portion of the driver. This is needed for XOrg (although not for XWayland), which will fall back to XOrg's mode-setting driver with GLAMOR if xserver-xorg-video-intel does not exist.

[tomtomme]

Ok I've got a beginner question here.
I thought that intel's graphics drivers/modules were actually included in the kernel source. So when I install something like xserver-xorg-video-intel, what am I actually installing? are these closed source blobs or something? I didn't think that intel graphics needed anything to be installed. Same with nouveau? What am I missing?
Thanks guys!

you do not have to install anything. xserver-bla-bla comes with your distro. It is part of the graphics stack of linux. This graphics stack consist of many parts, and only one is the kernel, another is mesa (for 3D), another is the Xserver etc. And all that is open source. See https://blogs.igalia.com/itoral/2014...raphics-stack/

[doublez13]

It's the xorg usermode portion of the driver. This is needed for XOrg (although not for XWayland), which will fall back to XOrg's mode-setting driver with GLAMOR if xserver-xorg-video-intel does not exist.

Ah ok that makes sense. Thanks guys!

[MadCatX]

Michael :

You wrote:



How did you enable DRI3 on xf86-video-modesetting? Xorg.0.log only mentions DRI2 here, see:

https://bugs.freedesktop.org/show_bug.cgi?id=98875

Does Xorg.0.log say DRI3 on your machine?

This is probably just a noise in the log. Try running

LIBGL_DEBUG=verbose glxinfo | grep DRI

for instance.

[shmerl]

How can I check if I'm using modesetting or Intel driver? I explicitly enabled DRI3 in a conf file, and can see it like this:

LIBGL_DEBUG=verbose glxinfo| grep DRI3
...
libGL: Using DRI3 for screen 0

But I can still see SNA mentioned on Xorg.0.log:

SNA initialized with Ivybridge (gen7, gt1) backend

[Sloth]

You can check the driver in /var/log/Xorg.0.log

On modesetting, libgl reports:

libGL: pci id for fd 4: 8086:191d, driver i965
libGL: OpenDriver: trying /usr/lib/xorg/modules/dri/tls/i965_dri.so
libGL: OpenDriver: trying /usr/lib/xorg/modules/dri/i965_dri.so
libGL: Using DRI3 for screen 0

[microcode]

Odd, I would have thought that gradients and texture mapping would be a perfect fit for fragment shaders.

[entropy]

Feeling a bit sorry for Chris Wilson.

He is a very skilled guy and he put lots of effort into SNA.
Hope he'll stay in Intel's open-source graphics division.

[balouba]

Meh, intel is not just much faster in the 2d tests. it obliterates modesetting. Its like 100-1000x faster...

For OpenGL tests I'm surprised: I run KDE Plasma right now with kwin/qt using OpenGL for UI acceleration: it is noticeably faster with the intel driver vs the modesetting driver, as in, while FPS seems about the same, UI interactions with the intel driver are noticeably lower latency (enough to be absolutely sure it's not placebo, so it's quite a bit faster).

Not sure why that is.

[patrakov]

One more bad thing about modesetting is that it brings ugly diagonal tearing on Haswell. Happens with a non-compositing window manager, happens with Compiz 0.8 (which I have to use because of color management via compicc add-on), doesn't happen with Cinnamon's Muffin.