- Dec 09, 2020
-
-
Ian Romanick authored
This fails mulitple ways on Mesa. The errant glBegin is included in the first display list , so an error is incorrectly generated by glCallList. The errant glReadPixels during compilation of the second display list does not generate an error.
-
Ian Romanick authored
-
Ian Romanick authored
In these cases the glBegin is executed, so the errors should be generated. This tests fails on Mesa.
-
Ian Romanick authored
Verify that rendering happens while compiling a display list with GL_COMPILE_AND_EXECUTE.
-
Ian Romanick authored
gl_PrimitiveID isn't added to core OpenGL until 3.2. There is some hardware that supports GL_EXT_gpu_shader4 but not OpenGL 3.2. Display lists are removed in OpenGL 3.2 Core Profile, and some drivers do not support Compatibility. Using GL_EXT_gpu_shader4 gives the widest possible driver / hardware coverage. Mesa fails points, lines, triangles, and quads. Reviewed-by:
Marek Olšák <marek.olsak@amd.com>
-
Ian Romanick authored
Both of the OpenGL 3.1 interaction subtests fail on Mesa. When OpenGL 3.1 support was added to Mesa, we had no intention of ever supporting GL_ARB_compatibility or Compatibility Profile. As a result, there was no reason to keep the client state for GL_NV_primitive restart and the server state for OpenGL 3.1 separate. I believe these added subtests are correct. There are other (as yet untested) interactions between the two sets of state that are unclear. For example, if GL_PRIMITIVE_RESTART_INDEX_NV is 1 and GL_PRIMITIVE_RESTART_INDEX is 2, and both enables are set, what is the effective restart index? To this end, I have submitted a spec issue to Khronos. https://github.com/KhronosGroup/OpenGL-API/issues/75 v2: Verify that glPrimitiveRestartIndexNV (and glPrimitiveRestartIndex) is not compiled into a display list. Reviewed-by:
Marek Olšák <marek.olsak@amd.com>
-
Yevhenii Kharchenko authored
Its purpose is to additionally validate depth and stencil data, that was read from PBO with glReadPixels(). Reading GL_DEPTH_COMPONENT, GL_STENCIL_INDEX, GL_DEPTH_STENCIL from following types: GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT24, GL_DEPTH_COMPONENT32F, GL_DEPTH24_STENCIL8, GL_DEPTH32F_STENCIL8, each combination as separate subtest. Tests: mesa/mesa#3775 Signed-off-by:
Yevhenii Kharchenko <yevhenii.kharchenko@globallogic.com> Reviewed-by:
Tapani Pälli <tapani.palli@intel.com> Part-of: <mesa/piglit!422>
-
- Dec 08, 2020
-
-
Ian Romanick authored
All of these tests currently pass on Iris. Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!427>
-
Ian Romanick authored
All tests pass on Mesa. v2: Don't call glGetError between glBegin and glEnd because it is not allowed. v3: Don't expect an error for glPrimitiveRestartIndex after a glBegin that is put in a display list. The glBegin is not "executed," so it does not affect the glPrimitiveRestartIndex call or the glGetError call. Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!427>
-
Ian Romanick authored
Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!427>
-
Ian Romanick authored
More tests will be added soon, so it makes sense to move them out of tests/general. Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!427>
-
Ian Romanick authored
Both tests pass on Ice Lake using Iris. Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!427>
-
Ian Romanick authored
Both tests pass on Ice Lake using Iris. Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!427>
-
Ian Romanick authored
Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!427>
-
Ian Romanick authored
The GL_ARB_shader_draw_parameters spec says: In non-MultiDraw* commands, the value of <gl_DrawIDARB> is always zero. This test contains two variations. The first variation attempts a bunch of non-MultiDraw* commands and verifies that gl_DrawIDARB is always zero. The second variation does the same thing using compatibility profile display lists. The display lists are constructed in a way that an implementation may coalesce the draws into a single operation that resembles a MultDraw* command. Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!427>
-
Ian Romanick authored
The extensions being tested only require OpenGL 3.1. The top-level OpenGL requirement of the tests is OpenGL 3.1. Commit 43aec6de ("arb_shader_draw_parameters: require GLSL 3.30") added a GLSL 3.30 requirement due to the use of layout("location=...") in the shaders. This is effectively an OpenGL 3.3 requirement. To increase (theoretical) test coverage, make these tests work on vanilla OpenGL 3.1 again. Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!427>
-
Ian Romanick authored
Found by inspection. Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!427>
-
- Dec 03, 2020
-
-
Andres Gomez authored
Signed-off-by:
Andres Gomez <agomez@igalia.com> Reviewed-by:
Dylan Baker <dylan@pnwbakers.com> Part-of: <mesa/piglit!424>
-
Andres Gomez authored
Previously, we were just accessing directly some private API. Signed-off-by:
Andres Gomez <agomez@igalia.com> Reviewed-by:
Dylan Baker <dylan@pnwbakers.com> Part-of: <mesa/piglit!424>
-
Andres Gomez authored
pid and time start/end was artificially added. Hence, when loading, we should remove those lines from the syserr output. Signed-off-by:
Andres Gomez <agomez@igalia.com> Reviewed-by:
Dylan Baker <dylan@pnwbakers.com> Part-of: <mesa/piglit!423>
-
Andres Gomez authored
v2: - Added fixes line. Fixes: 733dddbc ("framework/backends/junit.py: restore time from stderr") Signed-off-by:
Andres Gomez <agomez@igalia.com> Reviewed-by:
Dylan Baker <dylan@pnwbakers.com> Part-of: <mesa/piglit!423>
-
- Dec 02, 2020
-
-
Adam Jackson authored
The mode's already there, just not wired up for some reason.
-
Adam Jackson authored
Both of these extensions are legal under GLES2 and reuse the GL_READ_FRAMEBUFFER_BINDING token.
-
- Nov 28, 2020
-
-
Mark Kettenis authored
Currently the code uses exit, which runs the exit handlers installed by the parent. This causes issues since some of the Mesa drivers (such as radeonsi) install exit handlers to clean up the thread pool they created. However since the threads in the pool don't exist in the child this ends up calling pthread_join on non-existent threads which is undefined behaviour. Signed-off-by:
Mark Kettenis <kettenis@openbsd.org> Reviewed-by:
Michel Dänzer <mdaenzer@redhat.com> Part-of: <mesa/piglit!415>
-
- Nov 26, 2020
-
-
Daniel Stone authored
Pushing containers from Windows takes ages, like properly millions of years. Until we figure out why, just hope we don't need to rebuild the container too often, and set an outrageously long timeout to allow it to push. Signed-off-by:
Daniel Stone <daniels@collabora.com> Part-of: <mesa/piglit!378>
-
Daniel Stone authored
Rather than relying on global state for 'docker login' credentials, use a local file so we don't collide with other simultaneous builds. [cross-picked from mesa/mesa@a2808108] Signed-off-by:
Daniel Stone <daniels@collabora.com> Reviewed-by:
Erik Faye-Lund <erik.faye-lund@collabora.com> Part-of: <mesa/piglit!378>
-
- Nov 23, 2020
-
-
Mike Blumenkrantz authored
these tests should always generate at least 1 primitive, so don't allow drivers to sneak into a passing result if they fail to return at least 1 primitive here from the query (either from a draw failure or a query failure) Reviewed-by:
Samuel Iglesias Gonsálvez <siglesias@igalia.com> Part-of: <mesa/piglit!420>
-
Mike Blumenkrantz authored
Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!400>
-
Mike Blumenkrantz authored
this makes it easy to verify fp64 implementations using shader tests Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!400>
-
- Nov 21, 2020
-
-
Marcin Ślusarz authored
When hw spawns a lot of threads, the contention on a single atomic variable ("value") may be so high that the shader can take so much time that it can be mistakenly recognized as a GPU hang. This is what exactly happened when Intel Mesa 3D driver started to use SIMD32 mode (which increases the number of concurrent invocations) instead of SIMD16. As there's no point in testing values bigger than the mask array can hold, exit as soon as we are sure that the value will not be evaluated. v2 (Caio): * cast to matching types * use >= instead of > v3 (Marcin): * check value inside of the compare & swap loop, still hitting the original bug (https://bugs.freedesktop.org/show_bug.cgi?id=106774 ) v4 (Caio): * move even more code to the compare & swap loop v5 (Caio): * improve comment Closes mesa#3753. Signed-off-by:
Marcin Ślusarz <marcin.slusarz@intel.com> Tested-by: Clayton Craft <clayton.a.craft@intel.com> (v1) Reviewed-by:
Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com> Reviewed-by:
Ian Romanick <ian.d.romanick@intel.com> Part-of: <mesa/piglit!411>
-
- Nov 16, 2020
-
-
Andres Gomez authored
The included modules behavior may be modified by the toolchain file passed through the "CMAKE_TOOLCHAIN_FILE" variable, but this is not taken into account until the "project" command is invoked. References: https://gitlab.kitware.com/cmake/cmake/-/issues/21430 Signed-off-by:
Andres Gomez <agomez@igalia.com> Reviewed-by:
Jose Fonseca <jfonseca@vmware.com> Reviewed-by:
Erik Faye-Lund <erik.faye-lund@collabora.com> Reviewed-by:
Christian Gmeiner <christian.gmeiner@gmail.com> Reviewed-by:
Dylan Baker <dylan@pnwbakers.com> Part-of: <mesa/piglit!414>
-
Andres Gomez authored
After d03bca72 ("templates: fix loop iterating over value.images array"), the inclusion of images in the summary report was fixed but it also re-introduced a pre-existing problem: upon an OSError exception when running the X Test suite, the images elements may not include all the expected dict keywords. We now take care of that scenario, for backward compatibility, and make sure that all the keywords will exist in the future. Fixes: e1057c4a ("tests/xts.py: Add a tests file for the X Test suite.") Fixes: 5ef7f3b8 ("framework: Add xorg tests") Closes: mesa/piglit#43 Reported-by:
Michel Dänzer <mdaenzer@redhat.com> Signed-off-by:
Andres Gomez <agomez@igalia.com> Tested-by:
Michel Dänzer <mdaenzer@redhat.com> Reviewed-by:
Dylan Baker <dylan@pnwbakers.com> Part-of: <mesa/piglit!413>
-
- Nov 14, 2020
-
-
Nanley Chery authored
OpenGL 3.0 spec, section 4.2.3 "Clearing the Buffers": depth and stencil are the values to clear the depth and stencil buffers to, respectively. Clamping and type conversion for fixed-point depth buffers are performed in the same fashion as for ClearDepth. generate_simple_fbo generates a fixed-point depth buffer, so the clamping applies. Note that there's now some implicit communication happening between these two modules. Demonstrates that mesa/iris currently lacks clamping. Reviewed-by:
Lionel Landwerlin <lionel.g.landwerlin@intel.com> Part-of: <mesa/piglit!406>
-
Nanley Chery authored
OpenGL 3.0 spec, section 4.2.3 "Clearing the Buffers": If buffer is DEPTH, drawbuffer must be zero, and value points to the single depth value to clear the depth buffer to. Clamping and type conversion for fixed-point depth buffers are performed in the same fashion as for ClearDepth. generate_simple_fbo generates a fixed-point depth buffer, so the clamping applies. Note that there's now some implicit communication happening between these two modules. Demonstrates that mesa/iris currently lacks clamping. Reviewed-by:
Lionel Landwerlin <lionel.g.landwerlin@intel.com> Part-of: <mesa/piglit!406>
-
- Nov 13, 2020
-
-
Nanley Chery authored
Change red to something less than 1.0 in order to demonstrate a fast-clear bug in i965 and iris. Reviewed-by:
Tapani Pälli <tapani.palli@intel.com> Part-of: <mesa/piglit!292>
-
- Nov 12, 2020
-
-
Nanley Chery authored
Before this patch, this test cleared a GL_RGB32F texture to a NaN (0xffffffff), then checked for that exact bit pattern from the texelFetch instruction. It's possible that HW may return a different NaN encoding, so update the clear program to check for NaN instead. Enables iris to pass this test on gen11+. Closes: mesa/piglit#42 Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!412>
-
Andres Gomez authored
v2: - Improve resulting jobs names (Eric). Contributed by Michel Dänzer. Signed-off-by:
Andres Gomez <agomez@igalia.com> Suggested-by:
Michel Dänzer <mdaenzer@redhat.com> Reviewed-by: Eric Engestrom <eric@engestrom.ch> [v1] Reviewed-by:
Michel Dänzer <mdaenzer@redhat.com> Part-of: <mesa/piglit!394>
-
- Nov 11, 2020
-
-
Tapani Pälli authored
Currently framework requires external memory feature for all images, even when export did not happen. As example vk-image-display that requires export for depth attachment even though it is not required. This patch makes it possible for tests to set which buffers require export and thus makes it possible for drivers that would not support depth/stencil export to pass tests that do not require it. Signed-off-by:
Tapani Pälli <tapani.palli@intel.com> Reviewed-by:
Eleni Maria Stea <estea@igalia.com> Part-of: <mesa/piglit!397>
-
Erik Faye-Lund authored
This test calls fbo_formats_init, which in turn calls fbo_formats_init_test_set, which finally requires the GL_ARB_texture_env_combine extension. That extension makes no sense in core-contexts, where the fixed-function pipeline has been removed. So we end up bailing the entire test in this case. Running this in compatibility-context instead reveals that several mesa-drivers are broken, and crash when this test is run. Reviewed-by:
Marek Olšák <marek.olsak@amd.com> Part-of: <mesa/piglit!410>
-
Eleni Maria Stea authored
The purpose of the test is to reuse a Vulkan depth buffer from OpenGL. The test first draws something with Vulkan with the z buffer enabled in depth 0.33. The Vulkan z buffer is then imported in OpenGL, where it is used as the depth attachment of a framebuffer object. The color attachment of the same framebuffer object is first cleared to green. Then, it is used as render target to render a blue quad in depth 0.66. The final result must be: a blue quad with a smaller green quad in the middle, as the region that has depth 0.33 is hiding the blue quad. v2: * removed gl_target variable * used local variable vk_depth_format instead of global depth_format in the test v3: * fixed issue with draw: depth attachment is required otherwise the barrier won't be set and test won't pass on the GPUs that require it v4: * removed leftover comment from CMakeLists.gl.txt v5: * UINT 24/8 depth/stencil format is not supported on amdgpus, and so if the vendor is "AMD", we'd rather use a supported one. v6: * Replaced the v5 change with an iteration through all depth formats (splitted the test in subtests). v7: * Replaced left over code for AMD. v8: * Renamed all_formats array to depth_stencil_formats for clarity. Signed-off-by:
Eleni Maria Stea <estea@igalia.com>
-