4 Normally, no environment variables need to be set. Most of the
5 environment variables used by Mesa/Gallium are for debugging purposes,
6 but they can sometimes be useful for debugging end-user issues.
8 LibGL environment variables
9 ---------------------------
12 If defined debug information will be printed to stderr. If set to
13 ``verbose`` additional information will be printed.
14 ``LIBGL_DRIVERS_PATH``
15 colon-separated list of paths to search for DRI drivers
16 ``LIBGL_ALWAYS_INDIRECT``
17 if set to ``true``, forces an indirect rendering context/connection.
18 ``LIBGL_ALWAYS_SOFTWARE``
19 if set to ``true``, always use software rendering
20 ``LIBGL_NO_DRAWARRAYS``
21 if set to ``true``, do not use DrawArrays GLX protocol (for
24 print framerate to stdout based on the number of ``glXSwapBuffers``
26 ``LIBGL_DRI3_DISABLE``
27 disable DRI3 if set to ``true``.
29 Core Mesa environment variables
30 -------------------------------
33 if set, disables all assembly language optimizations
35 if set, disables Intel MMX optimizations
37 if set, disables AMD 3DNow! optimizations
39 if set, disables Intel SSE optimizations
41 if set to 1, error checking is disabled as per ``KHR_no_error``. This
42 will result in undefined behaviour for invalid use of the api, but
43 can reduce CPU use for apps that are known to be error free.
45 if set, error messages are printed to stderr. For example, if the
46 application generates a ``GL_INVALID_ENUM`` error, a corresponding
47 error message indicating where the error occurred, and possibly why,
48 will be printed to stderr. For release builds, ``MESA_DEBUG``
49 defaults to off (no debug output). ``MESA_DEBUG`` accepts the
50 following comma-separated list of named flags, which adds extra
51 behaviour to just set ``MESA_DEBUG=1``:
54 turn off debug messages. Only useful for debug builds.
56 flush after each drawing command
58 extra debug messages when a texture is incomplete
60 extra debug messages when a fbo is incomplete
62 create a debug context (see ``GLX_CONTEXT_DEBUG_BIT_ARB``) and
63 print error and performance messages to stderr (or
67 specifies a file name for logging all errors, warnings, etc., rather
70 if set, implement conventional texture env modes with fragment
71 programs (intended for developers only)
73 if set, implement conventional vertex transformation operations with
74 vertex programs (intended for developers only). Setting this variable
75 automatically sets the ``MESA_TEX_PROG`` variable as well.
76 ``MESA_EXTENSION_OVERRIDE``
77 can be used to enable/disable extensions. A value such as
78 ``GL_EXT_foo -GL_EXT_bar`` will enable the ``GL_EXT_foo`` extension
79 and disable the ``GL_EXT_bar`` extension.
80 ``MESA_EXTENSION_MAX_YEAR``
81 The ``GL_EXTENSIONS`` string returned by Mesa is sorted by extension
82 year. If this variable is set to year X, only extensions defined on
83 or before year X will be reported. This is to work-around a bug in
84 some games where the extension string is copied into a fixed-size
85 buffer without truncating. If the extension string is too long, the
86 buffer overrun can cause the game to crash. This is a work-around for
88 ``MESA_GL_VERSION_OVERRIDE``
89 changes the value returned by ``glGetString(GL_VERSION)`` and
90 possibly the GL API type.
92 - The format should be ``MAJOR.MINOR[FC|COMPAT]``
93 - ``FC`` is an optional suffix that indicates a forward compatible
94 context. This is only valid for versions >= 3.0.
95 - ``COMPAT`` is an optional suffix that indicates a compatibility
96 context or ``GL_ARB_compatibility`` support. This is only valid
98 - GL versions <= 3.0 are set to a compatibility (non-Core) profile
99 - GL versions = 3.1, depending on the driver, it may or may not have
100 the ``ARB_compatibility`` extension enabled.
101 - GL versions >= 3.2 are set to a Core profile
105 select a compatibility (non-Core) profile with GL version 2.1.
107 select a compatibility (non-Core) profile with GL version 3.0.
109 select a Core+Forward Compatible profile with GL version 3.0.
111 select GL version 3.1 with ``GL_ARB_compatibility`` enabled per
114 select GL version 3.1 with forward compatibility and
115 ``GL_ARB_compatibility`` disabled.
117 select GL version 3.1 with ``GL_ARB_compatibility`` enabled.
119 override GL version to X.Y without changing the profile.
121 select a Core+Forward Compatible profile with GL version X.Y.
123 select a Compatibility profile with GL version X.Y.
125 - Mesa may not really implement all the features of the given
126 version. (for developers only)
128 ``MESA_GLES_VERSION_OVERRIDE``
129 changes the value returned by ``glGetString(GL_VERSION)`` for OpenGL
132 - The format should be ``MAJOR.MINOR``
133 - Examples: ``2.0``, ``3.0``, ``3.1``
134 - Mesa may not really implement all the features of the given
135 version. (for developers only)
137 ``MESA_GLSL_VERSION_OVERRIDE``
138 changes the value returned by
139 ``glGetString(GL_SHADING_LANGUAGE_VERSION)``. Valid values are
140 integers, such as ``130``. Mesa will not really implement all the
141 features of the given language version if it's higher than what's
142 normally reported. (for developers only)
143 ``MESA_GLSL_CACHE_DISABLE``
144 if set to ``true``, disables the GLSL shader cache
145 ``MESA_GLSL_CACHE_MAX_SIZE``
146 if set, determines the maximum size of the on-disk cache of compiled
147 GLSL programs. Should be set to a number optionally followed by
148 ``K``, ``M``, or ``G`` to specify a size in kilobytes, megabytes, or
149 gigabytes. By default, gigabytes will be assumed. And if unset, a
150 maximum size of 1GB will be used.
154 A separate cache might be created for each architecture that Mesa is
155 installed for on your system. For example under the default settings
156 you may end up with a 1GB cache for x86_64 and another 1GB cache for
159 ``MESA_GLSL_CACHE_DIR``
160 if set, determines the directory to be used for the on-disk cache of
161 compiled GLSL programs. If this variable is not set, then the cache
162 will be stored in ``$XDG_CACHE_HOME/mesa_shader_cache`` (if that
163 variable is set), or else within ``.cache/mesa_shader_cache`` within
164 the user's home directory.
166 :ref:`shading language compiler options <envvars>`
167 ``MESA_NO_MINMAX_CACHE``
168 when set, the minmax index cache is globally disabled.
169 ``MESA_SHADER_CAPTURE_PATH``
170 see :ref:`Capturing Shaders <capture>`
171 ``MESA_SHADER_DUMP_PATH`` and ``MESA_SHADER_READ_PATH``
172 see :ref:`Experimenting with Shader
173 Replacements <replacement>`
174 ``MESA_VK_VERSION_OVERRIDE``
175 changes the Vulkan physical device version as returned in
176 ``VkPhysicalDeviceProperties::apiVersion``.
178 - The format should be ``MAJOR.MINOR[.PATCH]``
179 - This will not let you force a version higher than the driver's
180 instance version as advertised by ``vkEnumerateInstanceVersion``
181 - This can be very useful for debugging but some features may not be
182 implemented correctly. (For developers only)
184 NIR passes enviroment variables
185 -------------------------------
187 The following are only applicable for drivers that uses NIR, as they
188 modify the behaviour for the common NIR_PASS and NIR_PASS_V macros, that
189 wrap calls to NIR lowering/optimizations.
192 If defined, the resulting NIR shader will be printed out at each
193 succesful NIR lowering/optimization call.
195 If defined, cloning a NIR shader would be tested at each succesful
196 NIR lowering/optimization call.
197 ``NIR_TEST_SERIALIZE``
198 If defined, serialize and deserialize a NIR shader would be tested at
199 each succesful NIR lowering/optimization call.
201 Mesa Xlib driver environment variables
202 --------------------------------------
204 The following are only applicable to the Mesa Xlib software driver. See
205 the :doc:`Xlib software driver page <xlibdriver>` for details.
208 specifies the X visual and depth for RGB mode
210 specifies the X visual and depth for CI mode
212 specifies how to implement the back color buffer, either ``pixmap``
215 gamma correction coefficients for red, green, blue channels
217 enable synchronous X behavior (for debugging only)
218 ``MESA_GLX_FORCE_CI``
219 if set, force GLX to treat 8bpp visuals as CI visuals
220 ``MESA_GLX_FORCE_ALPHA``
221 if set, forces RGB windows to have an alpha channel.
222 ``MESA_GLX_DEPTH_BITS``
223 specifies default number of bits for depth buffer.
224 ``MESA_GLX_ALPHA_BITS``
225 specifies default number of bits for alpha channel.
227 i945/i965 driver environment variables (non-Gallium)
228 ----------------------------------------------------
231 if set to 1, prevents batches from being submitted to the hardware.
232 This is useful for debugging hangs, etc.
234 a comma-separated list of named flags, which do various things:
237 annotate IR in assembly dumps
239 dump batches into an AUB trace for use with simulation tools
241 emit batch information
243 emit messages about blit operations
245 emit messages about the blorp operations (blits & clears)
247 emit messages about buffer objects
249 emit messages about the clip unit (for old gens, includes the CLIP
254 dump shader assembly for compute shaders
256 generate compute shader SIMD32 programs even if workgroup size
257 doesn't exceed the SIMD16 limit
259 emit messages about the DRI interface
261 emit messages about framebuffers
263 dump shader assembly for fragment shaders
265 dump shader assembly for geometry shaders
267 print instruction hex dump with the disassembly
269 emit messages about the new L3 state during transitions
271 emit messages about miptrees
273 don't generate SIMD8 fragment shader
275 suppress generation of 16-wide fragment shaders. useful for
276 debugging broken shaders
278 disable instruction compaction
280 suppress generation of dual-object geometry shader code
284 disable single sampled render buffer compression
286 dump shader assembly to files at each optimization pass and
287 iteration that make progress
289 emit messages about performance issues
291 emit messages about ``AMD_performance_monitor``
293 emit messages about pixel operations
295 emit messages about drawing primitives
297 mark all state dirty on each draw call
299 emit messages about the strips & fans unit (for old gens, includes
302 record how much GPU time is spent in each shader
304 force spilling of all registers in the scalar backend (useful to
307 force spilling of all registers in the vec4 backend (useful to
310 emit messages about state flag tracking
312 emit batchbuffer usage statistics
314 after sending each batch, emit a message and wait for that batch
317 dump shader assembly for tessellation control shaders
319 dump shader assembly for tessellation evaluation shaders
321 emit messages about textures.
323 emit messages about URB setup
325 emit messages about vertex assembly
327 dump shader assembly for vertex shaders
329 ``INTEL_SCALAR_VS`` (or ``TCS``, ``TES``, ``GS``)
330 force scalar/vec4 mode for a shader stage (Gen8-9 only)
331 ``INTEL_PRECISE_TRIG``
332 if set to 1, true or yes, then the driver prefers accuracy over
333 performance in trig functions.
335 Radeon driver environment variables (radeon, r200, and r300g)
336 -------------------------------------------------------------
339 if set, disable hardware-accelerated Transform/Clip/Lighting.
341 EGL environment variables
342 -------------------------
344 Mesa EGL supports different sets of environment variables. See the
345 :doc:`Mesa EGL <egl>` page for the details.
347 Gallium environment variables
348 -----------------------------
351 draws various information on the screen, like framerate, cpu load,
352 driver statistics, performance counters, etc. Set
353 ``GALLIUM_HUD=help`` and run e.g. ``glxgears`` for more info.
354 ``GALLIUM_HUD_PERIOD``
355 sets the hud update rate in seconds (float). Use zero to update every
356 frame. The default period is 1/2 second.
357 ``GALLIUM_HUD_VISIBLE``
358 control default visibility, defaults to true.
359 ``GALLIUM_HUD_TOGGLE_SIGNAL``
360 toggle visibility via user specified signal. Especially useful to
361 toggle hud at specific points of application and disable for
362 unencumbered viewing the rest of the time. For example, set
363 ``GALLIUM_HUD_VISIBLE`` to ``false`` and
364 ``GALLIUM_HUD_TOGGLE_SIGNAL`` to ``10`` (``SIGUSR1``). Use
365 ``kill -10 <pid>`` to toggle the hud as desired.
366 ``GALLIUM_HUD_SCALE``
367 Scale hud by an integer factor, for high DPI displays. Default is 1.
368 ``GALLIUM_HUD_DUMP_DIR``
369 specifies a directory for writing the displayed hud values into
372 useful in combination with ``LIBGL_ALWAYS_SOFTWARE=true`` for
373 choosing one of the software renderers ``softpipe``, ``llvmpipe`` or
376 specifies a file for logging all errors, warnings, etc. rather than
378 ``GALLIUM_PRINT_OPTIONS``
379 if non-zero, print all the Gallium environment variables which are
380 used, and their current values.
382 if non-zero, print information about the CPU on start-up
383 ``TGSI_PRINT_SANITY``
384 if set, do extra sanity checking on TGSI shaders and print any errors
391 if set to zero, the draw module will not use LLVM to execute shaders,
394 controls debug output from the Mesa/Gallium state tracker. Setting to
395 ``tgsi``, for example, will print all the TGSI shaders. See
396 ``src/mesa/state_tracker/st_debug.c`` for other options.
398 Clover environment variables
399 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
401 ``CLOVER_EXTRA_BUILD_OPTIONS``
402 allows specifying additional compiler and linker options. Specified
403 options are appended after the options set by the OpenCL program in
405 ``CLOVER_EXTRA_COMPILE_OPTIONS``
406 allows specifying additional compiler options. Specified options are
407 appended after the options set by the OpenCL program in
408 ``clCompileProgram``.
409 ``CLOVER_EXTRA_LINK_OPTIONS``
410 allows specifying additional linker options. Specified options are
411 appended after the options set by the OpenCL program in
414 Softpipe driver environment variables
415 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
418 if set, the softpipe driver will print fragment shaders to stderr
420 if set, the softpipe driver will print geometry shaders to stderr
422 if set, rasterization is no-op'd. For profiling purposes.
423 ``SOFTPIPE_USE_LLVM``
424 if set, the softpipe driver will try to use LLVM JIT for vertex
427 LLVMpipe driver environment variables
428 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
431 if set LLVMpipe will no-op rasterization
433 a comma-separated list of debug options is accepted. See the source
436 a comma-separated list of options to selectively no-op various parts
437 of the driver. See the source code for details.
439 an integer indicating how many threads to use for rendering. Zero
440 turns off threading completely. The default value is the number of
443 VMware SVGA driver environment variables
444 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
447 force use of software vertex transformation
449 don't allow software vertex transformation fallbacks (will often
450 result in incorrect rendering).
452 for dumping shaders, constant buffers, etc. See the code for details.
453 ``SVGA_EXTRA_LOGGING``
454 if set, enables extra logging to the ``vmware.log`` file, such as the
455 OpenGL program's name and command line arguments.
457 if set, disables logging to the ``vmware.log`` file. This is useful
458 when using Valgrind because it otherwise crashes when initializing
459 the host log feature.
461 See the driver code for other, lesser-used variables.
463 WGL environment variables
464 ~~~~~~~~~~~~~~~~~~~~~~~~~
466 ``WGL_SWAP_INTERVAL``
467 to set a swap interval, equivalent to calling
468 ``wglSwapIntervalEXT()`` in an application. If this environment
469 variable is set, application calls to ``wglSwapIntervalEXT()`` will
472 VA-API environment variables
473 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
475 ``VAAPI_MPEG4_ENABLED``
476 enable MPEG4 for VA-API, disabled by default.
478 VC4 driver environment variables
479 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
482 a comma-separated list of named flags, which do various things:
485 dump command list during creation
487 dump generated QPU instructions
489 dump QPU IR during program compile
491 dump NIR during program compile
493 dump TGSI during program compile
495 dump program compile information for shader-db analysis
497 print during performance-related events
499 skip actual hardware execution of commands
501 flush after each draw call
503 wait for finish after each flush
505 write a GPU command stream trace file (VC4 simulator only)
507 RADV driver environment variables
508 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
511 a comma-separated list of named flags, which do various things:
514 force all allocated buffers to be referenced in submissions
516 enable all device/instance entrypoints
518 validate the LLVM IR before LLVM compiles the shader
520 display more info about errors
522 show GPU-related information
524 dump internal meta shaders
526 disable primitive binning
528 disable shaders cache
530 disable compute queue
532 disable Delta Color Compression (DCC) on images
534 do not check OOB access for dynamic descriptors
536 disable fast color/depthstencil clears
538 disable HIZ for depthstencil images
540 disable directly recording command buffers in GPU-visible memory
542 disable LLVM SILoadStoreOptimizer pass
544 disable memory shaders cache
546 disable NGG for GFX10+
548 disable out-of-order rasterization
550 disable shader ballot
552 disable LLVM threaded compilation
554 dump LLVM IR before any optimizations
558 dump shader statistics
562 display info at startup
564 synchronize shaders after all draws/dispatches
566 check for VM memory faults via dmesg
568 initialize all memory allocated in VRAM as zero
570 ``RADV_FORCE_FAMILY``
571 create a null device to compile shaders without a AMD GPU (eg.
574 a comma-separated list of named flags, which do various things:
577 enable ACO experimental compiler
579 enable the global BO list
581 enable wave32 for compute shaders (GFX10+)
583 enable DCC for MSAA images
587 enable wave32 for vertex/tess/geometry shaders (GFX10+)
591 enable wave32 for pixel shaders (GFX10+)
595 enable TC-compat cmask for MSAA images
597 ``RADV_SECURE_COMPILE_THREADS``
598 maximum number of secure compile threads (up to 32)
600 force anisotropy filter (up to 16)
602 generate cmdbuffer tracefiles when a GPU hang is detected
604 a comma-separated list of named flags, which do various things:
607 validate the ACO IR at various points of compilation (enabled by
608 default for debug/debugoptimized builds)
610 validate register assignment of ACO IR and catches many RA bugs
612 abort on some suboptimal code generation
614 radeonsi driver environment variables
615 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
618 a comma-separated list of named flags, which do various things:
624 Disable SDMA image copies
626 Clear VRAM allocations.
630 Disable DCC fast clear.
632 Disable separate DCC on the main framebuffer
642 Disable MSAA compression
650 Print driver information
656 Print virtual addresses when creating resources
662 Print geometry shaders
664 Print tessellation control shaders
666 Print tessellation evaluation shaders
668 Print compute shaders
670 Don't print the LLVM IR
672 Don't print NIR when printing shaders
674 Don't print disassembled shaders
676 Print the LLVM IR before initial optimizations
678 Enable LLVM global instruction selector.
680 Use Wave32 for vertex, tessellation, and geometry shaders.
682 Use Wave32 for pixel shaders.
684 Use Wave32 for computes shaders.
686 Use Wave64 for vertex, tessellation, and geometry shaders.
688 Use Wave64 for pixel shaders.
690 Use Wave64 for computes shaders.
692 Enable additional sanity checks on shader IR
694 Use old-style monolithic shaders compiled on demand
696 Disable compiling optimized shader variants.
698 Use SDMA for all operations when possible.
700 Disable GTT write combining
702 Check VM faults and dump debug info.
704 Force VMID reservation per context.
706 Disable graphics. Only multimedia compute paths can be used.
708 Disable NGG and use the legacy pipeline.
710 Always use NGG culling even when it can hurt.
714 Always enable the primitive discard compute shader.
716 Enable the primitive discard compute shader for large draw calls.
718 Disable the primitive discard compute shader.
720 Program WD/IA to switch on end-of-packet.
722 Disable out-of-order rasterization
728 Other Gallium drivers have their own environment variables. These may
729 change frequently so the source code should be consulted for details.