Chapter 11. Specifying OpenGL Environment Variable Settings

Full scene antialiasing

Antialiasing is a technique used to smooth the edges of objects in a scene to reduce the jagged "stairstep" effect that sometimes appears. By setting the appropriate environment variable, you can enable full-scene antialiasing in any OpenGL application on these GPUs.

Several antialiasing methods are available and you can select between them by setting the __GL_FSAA_MODE environment variable appropriately. Note that increasing the number of samples taken during FSAA rendering may decrease performance.

To see the available values for __GL_FSAA_MODE along with their descriptions, run:

    nvidia-settings --query=fsaa --verbose

The __GL_FSAA_MODE environment variable uses the same integer values that are used to configure FSAA through nvidia-settings and the NV-CONTROL X extension. In other words, these two commands are equivalent:

    export __GL_FSAA_MODE=5

    nvidia-settings --assign FSAA=5

Note that there are three FSAA related configuration attributes (FSAA, FSAAAppControlled and FSAAAppEnhanced) which together determine how a GL application will behave. If FSAAAppControlled is 1, the FSAA specified through nvidia-settings will be ignored, in favor of what the application requests through FBConfig selection. If FSAAAppControlled is 0 but FSAAAppEnhanced is 1, then the FSAA value specified through nvidia-settings will only be applied if the application selected a multisample FBConfig.

Therefore, to be completely correct, the nvidia-settings command line to unconditionally assign FSAA should be:

    nvidia-settings --assign FSAA=5 --assign FSAAAppControlled=0 --assign FSAAAppEnhanced=0

The driver may not be able to support a particular FSAA mode for a given application due to video or system memory limitations. In that case, the driver will silently fall back to a less demanding FSAA mode.

Anisotropic texture filtering

Automatic anisotropic texture filtering can be enabled by setting the environment variable __GL_LOG_MAX_ANISO. The possible values are:

__GL_LOG_MAX_ANISO Filtering Type
0 No anisotropic filtering
1 2x anisotropic filtering
2 4x anisotropic filtering
3 8x anisotropic filtering
4 16x anisotropic filtering

4x and greater are only available on GeForce3 or newer GPUs; 16x is only available on GeForce 6800 or newer GPUs.

Vblank syncing

The __GL_SYNC_TO_VBLANK (boolean) environment variable can be used to control whether swaps are synchronized to a display device's vertical refresh.

When sync to vblank is enabled with TwinView, OpenGL can only sync to one of the display devices; this may cause tearing corruption on the display device to which OpenGL is not syncing. You can use the environment variable __GL_SYNC_DISPLAY_DEVICE to specify to which display device OpenGL should sync. You should set this environment variable to the name of a display device; for example "CRT-1". Look for the line "Connected display device(s):" in your X log file for a list of the display devices present and their names. You may also find it useful to review Chapter 12, Configuring Multiple Display Devices on One X Screen "Configuring Twinview" and the section on Ensuring Identical Mode Timings in Chapter 19, Programming Modes.

Controlling the sorting of OpenGL FBConfigs

The NVIDIA GLX implementation sorts FBConfigs returned by glXChooseFBConfig() as described in the GLX specification. To disable this behavior set __GL_SORT_FBCONFIGS to 0 (zero), then FBConfigs will be returned in the order they were received from the X server. To examine the order in which FBConfigs are returned by the X server run:

nvidia-settings --glxinfo

This option may be be useful to work around problems in which applications pick an unexpected FBConfig.

OpenGL yield behavior

There are several cases where the NVIDIA OpenGL driver needs to wait for external state to change before continuing. To avoid consuming too much CPU time in these cases, the driver will sometimes yield so the kernel can schedule other processes to run while the driver waits. For example, when waiting for free space in a command buffer, if the free space has not become available after a certain number of iterations, the driver will yield before it continues to loop.

By default, the driver calls sched_yield() to do this. However, this can cause the calling process to be scheduled out for a relatively long period of time if there are other, same-priority processes competing for time on the CPU. One example of this is when an OpenGL-based composite manager is moving and repainting a window and the X server is trying to update the window as it moves, which are both CPU-intensive operations.

You can use the __GL_YIELD environment variable to work around these scheduling problems. This variable allows the user to specify what the driver should do when it wants to yield. The possible values are:

__GL_YIELD Behavior
<unset> By default, OpenGL will call sched_yield() to yield.
"NOTHING" OpenGL will never yield.
"USLEEP" OpenGL will call usleep(0) to yield.

Controlling which OpenGL FBConfigs are available

The NVIDIA GLX implementation will hide FBConfigs that are associated with a 32-bit ARGB visual when the XLIB_SKIP_ARGB_VISUALS environment variable is defined. This matches the behavior of libX11, which will hide those visuals from XGetVisualInfo and XMatchVisualInfo. This environment variable is useful when applications are confused by the presence of these FBConfigs.

Using Unofficial GLX protocol

By default, the NVIDIA GLX implementation will not expose GLX protocol for GL commands if the protocol is not considered complete. Protocol could be considered incomplete for a number of reasons. The implementation could still be under development and contain known bugs, or the protocol specification itself could be under development or going through review. If users would like to test the client-side portion of such protocol when using indirect rendering, they can set the __GL_ALLOW_UNOFFICIAL_PROTOCOL environment variable to a non-zero value before starting their GLX application. When an NVIDIA GLX server is used, the related X Config option AllowUnofficialGLXProtocol will need to be set as well to enable support in the server.

Overriding driver detection of SELinux policy booleans

On Linux, the NVIDIA GLX implementation will attempt to detect whether SELinux is enabled and modify its behavior to respect SELinux policy. By default, the driver adheres to SELinux policy boolean settings at the beginning of a client process's execution; due to shared library limitations, these settings remain fixed throughout the lifetime of the driver instance. Additionally, the driver will adhere to policy boolean settings regardless of whether SELinux is running in permissive mode or enforcing mode. The __GL_SELINUX_BOOLEANS environment variable allows the user to override driver detection of specified SELinux booleans so the driver acts as if these booleans were set or unset. This allows the user, for example, to run the driver under a more restrictive policy than specified by SELinux, or to work around problems when running the driver under SELinux while operating in permissive mode.

__GL_SELINUX_BOOLEANS should be set to a comma-separated list of key/value pairs:


Valid keys are any SELinux booleans specified by "getsebool -a", and valid values are 1, true, yes, or on to enable the boolean, and 0, false, no, or off to disable it. There should be no whitespace between any key, value, or delimiter. Currently, the driver only uses the allow_execmem boolean to determine whether it can apply optimizations that use executable memory.

Limiting heap allocations in the OpenGL driver

The NVIDIA OpenGL implementation normally does not enforce limits on dynamic system memory allocations (i.e., memory allocated by the driver from the C library via the malloc(3) memory allocation package). The __GL_HEAP_ALLOC_LIMIT environment variable enables the user to specify a per-process heap allocation limit for as long as libGL is loaded in the application.

__GL_HEAP_ALLOC_LIMIT is specified in the form BYTES SUFFIX, where BYTES is a nonnegative integer and SUFFIX is an optional multiplicative suffix: kB = 1000, k = 1024, MB = 1000*1000, M = 1024*1024, GB = 1000*1000*1000, and G = 1024*1024*1024. SUFFIX is not case-sensitive. For example, to specify a heap allocation limit of 20 megabytes:


If SUFFIX is not specified, the limit is assumed to be given in bytes. The minimum heap allocation limit is 12 MB. If a lower limit is specified, the limit is clamped to the minimum.

The GNU C library provides several hooks that may be used by applications to modify the behavior of malloc(3), realloc(3), and free(3). In addition, an application or library may specify allocation symbols that the driver will use in place of those exported by libc. Heap allocation tracking is incompatible with these features, and the driver will disable the heap allocation limit if it detects that they are in use.

WARNING: Enforcing a limit on heap allocations may cause unintended behavior and lead to application crashes, data corruption, and system instability. ENABLE AT YOUR OWN RISK.

OpenGL Shader Disk Cache

The NVIDIA OpenGL driver utilizes a shader disk cache. This optimization benefits some applications, by reusing shader binaries instead of compiling them repeatedly. The related environment variables __GL_SHADER_DISK_CACHE and __GL_SHADER_DISK_CACHE_PATH, as well as the GLShaderDiskCache X configuration option, allow fine-grained configuration of the shader cache behavior. The shader disk cache:

  1. is always disabled for indirect rendering

  2. is always disabled for setuid and setgid binaries

  3. by default, is disabled for direct rendering when the OpenGL application is run as the root user

  4. by default, is enabled for direct rendering when the OpenGL application is run as a non-root user

The GLShaderDiskCache X configuration option forcibly enables or disables the shader disk cache, for direct rendering as a non-root user.

By default, caches are stored in $HOME/.nv/GLCache. Caches are persistent across runs of an application. Cached shader binaries are specific to each driver version; changing driver versions will cause binaries to be recompiled.

The following environment variables configure shader disk cache behavior, and override the GLShaderDiskCache configuration option:

Environment Variable Description
__GL_SHADER_DISK_CACHE (boolean) Enables or disables the shader cache for direct rendering.
__GL_SHADER_DISK_CACHE_PATH (string) Enables configuration of where shader caches are stored on disk.

Threaded Optimizations

The NVIDIA OpenGL driver supports offloading its CPU computation to a worker thread. These optimizations typically benefit CPU-intensive applications, but might cause a decrease of performance in applications that heavily rely on synchronous OpenGL calls such as glGet*. Because of this, they are currently disabled by default.

Setting the __GL_THREADED_OPTIMIZATIONS environment variable to "1" before loading the NVIDIA OpenGL driver library will enable these optimizations for the lifetime of the application.

Please note that these optimizations can currently only be enabled if the target application dynamically links against pthreads. If this isn't the case, the dynamic loader can be instructed to do so at runtime by setting the LD_PRELOAD environment variable to include the pthreads library.

Additionally, these optimizations require Xlib to function in thread-safe mode. The NVIDIA OpenGL driver will automatically attempt to enable Xlib thread-safe mode if needed. However, it might not be possible in some situations, such as when the NVIDIA OpenGL driver library is dynamically loaded after Xlib has been loaded and initialized. If that is the case, threaded optimizations will stay disabled unless the application is modified to call XInitThreads() before initializing Xlib or to link directly against the NVIDIA OpenGL driver library. Alternatively, using the LD_PRELOAD environment variable to include the NVIDIA OpenGL driver library should also achieve the desired result.

This sample command line should address all the limitations described above and allow the target application to run with NVIDIA threaded optimizations: