Xorg-7 Testing and Configuration

Checking the Direct Rendering Infrastructure (DRI) Installation

DRI is a framework for allowing software to access graphics hardware in a safe and efficient manner. It is installed in X by default (using Mesa) if you have a supported video card.

To check if DRI drivers are installed properly, check the log file /var/log/Xorg.0.log for statements such as:

(II) intel(0): direct rendering: DRI2 Enabled

or

(II) NOUVEAU(0): Loaded DRI module

Note

DRI configuration may differ if you are using alternate drivers, such as those from NVIDIA or AMD.

Although all users can use software acceleration, any hardware acceleration (DRI2) is only available to root and members of the video group, but ConsoleKit2 takes care of adding any logged in user to the user ACL's of /dev/dri/card*, the special file(s) allowing access to hardware acceleration.

If your driver is supported and ConsoleKit2 is not installed, add any users that might use X to the video group:

usermod -a -G video <username>

Another way to determine if DRI is working properly is to use one of the two optionally installed OpenGL demo programs in Mesa-18.1.6. From an X terminal, run glxinfo and look for the phrase:

name of display: :0
display: :0  screen: 0
direct rendering: Yes

If direct rendering is enabled, you can add verbosity by running LIBGL_DEBUG=verbose glxinfo. This will show the drivers, device nodes and files used by the DRI system.

To confirm that DRI2 hardware acceleration is working, you can (still in the X terminal) run the command glxinfo | egrep "(OpenGL vendor|OpenGL renderer|OpenGL version)". If that reports something other than Software Rasterizer then you have working acceleration for the user who ran the command.

If your hardware does not have any DRI2 driver available, it will use a Software Rasterizer for Direct Rendering. In such cases, you can use a new, LLVM-accelerated, Software Rasterizer called LLVMPipe. In order to build LLVMPipe just make sure that LLVM-6.0.1 is present at Mesa build time. Note that all decoding is done on the CPU instead of the GPU, so the display will run slower than with hardware acceleration. To check if you are using LLVMpipe, review the output of the glxinfo command above. An example of the output using the Software Rasterizer is shown below:

OpenGL vendor string: VMware, Inc.
OpenGL renderer string: Gallium 0.4 on llvmpipe (LLVM 3.5, 256 bits)
OpenGL version string: 3.0 Mesa 10.4.5

You can also force LLVMPipe by exporting the LIBGL_ALWAYS_SOFTWARE=1 environment variable when starting Xorg.

Again, if you have built the Mesa OpenGL demos, you can also run the test program glxgears. This program brings up a window with three gears turning. The X terminal will display how many frames were drawn every five seconds, so this will give a rough benchmark. The window is scalable, and the frames drawn per second is highly dependent on the size of the window. On some hardware, glxgears will run synchronized with the vertical refresh signal and the frame rate will be approximately the same as the monitor refresh rate.

Hybrid Graphics

Hybrid Graphics is still in experimental state for Linux. Xorg Developers have developed a technology called PRIME that can be used for switching between integrated and muxless discrete GPU at will. Automatic switching is not possible at the moment.

In order to use PRIME for GPU switching, make sure that you are using Linux Kernel 3.4 or later (recommended). You will need latest DRI and DDX drivers for your hardware and Xorg Server 1.13 or later.

Xorg Server should load both GPU drivers automaticaly. You can check that by running:

xrandr --listproviders

There should be two (or more) providers listed, for example:

Providers: number : 2
Provider 0: id: 0x7d cap: 0xb, Source Output, Sink Output, Sink Offload crtcs: 3 outputs: 4 associated providers: 1 name:Intel
Provider 1: id: 0x56 cap: 0xf, Source Output, Sink Output, Source Offload, Sink Offload crtcs: 6 outputs: 1 associated providers: 1 name:radeon

In order to be able to run a GLX application on a discrete GPU, you will need to run the following command, where <provider> is the more powerful discrete card, and <sink> is the card which has a display connected:

xrandr --setprovideroffloadsink <provider> <sink>

Note

With newer Xorg drivers, such as modesetting or intel, which are DRI3 capable, the above command is no longer necessary. It does no harm however.

Then, you will need to export the DRI_PRIME=1 environment variable each time you want the powerful GPU to be used. For example,

DRI_PRIME=1 glxinfo | egrep "(OpenGL vendor|OpenGL renderer|OpenGL version)"

will show OpenGL vendor, renderer and version for the discrete GPU.

If the last command reports same OpenGL renderer with and without DRI_PRIME=1, you will need to check your installation.

Setting up Xorg Devices

For most hardware configurations, modern Xorg will automatically get the server configuration correct without any user intervention. There are, however, some cases where auto-configuration will be incorrect. Following are some example manual configuration items that may be of use in these instances.

cat > /etc/X11/xorg.conf.d/xkb-defaults.conf << "EOF"
Section "InputClass"
    Identifier "XKB Defaults"
    MatchIsKeyboard "yes"
    Option "XkbLayout" "fr"
    Option "XkbOptions" "terminate:ctrl_alt_bksp"
EndSection
EOF

cat > /etc/X11/xorg.conf.d/videocard-0.conf << "EOF"
Section "Device"
    Identifier  "Videocard0"
    Driver      "radeon"
    VendorName  "Videocard vendor"
    BoardName   "ATI Radeon 7500"
    Option      "NoAccel" "true"
EndSection
EOF

cat > /etc/X11/xorg.conf.d/server-layout.conf << "EOF"
Section "ServerLayout"
    Identifier     "DefaultLayout"
    Screen      0  "Screen0" 0 0
    Screen      1  "Screen1" LeftOf "Screen0"
    Option         "Xinerama"
EndSection
EOF