Matthew Netsch [Mon, 14 Mar 2022 17:17:55 +0000 (17:17 +0000)]
Merge vk-gl-cts/opengl-cts-4.6.2 into vk-gl-cts/main
Change-Id: I6a1b91fafab2a290283303d75b4f8048af2113b8
Matthew Netsch [Mon, 14 Mar 2022 16:51:00 +0000 (16:51 +0000)]
Merge vk-gl-cts/opengl-cts-4.6.1 into vk-gl-cts/opengl-cts-4.6.2
Change-Id: I27e9df2eefc1ca63153c2fb10baf6b4118feaf4e
Matthew Netsch [Mon, 14 Mar 2022 16:00:35 +0000 (16:00 +0000)]
Merge vk-gl-cts/opengl-es-cts-3.2.8 into vk-gl-cts/main
Change-Id: Idfcd3157d21933d06a69356a3ce3c28f84fccb3c
Matthew Netsch [Mon, 14 Mar 2022 14:08:56 +0000 (14:08 +0000)]
Merge vk-gl-cts/opengl-es-cts-3.2.7 into vk-gl-cts/opengl-es-cts-3.2.8
Change-Id: I9dbfb6be27277508e01e4d79583afc5f970353e2
Matthew Netsch [Thu, 10 Mar 2022 21:58:48 +0000 (21:58 +0000)]
Merge vk-gl-cts/opengl-es-cts-3.2.6 into vk-gl-cts/opengl-es-cts-3.2.7
Change-Id: I216b8bf396d0ff5d4ceaa890d5b4dd9283c73040
Ricardo Garcia [Thu, 3 Mar 2022 10:52:58 +0000 (11:52 +0100)]
Avoid logging to stdout in buffer mem requirements tests
Redirect informative messages to the output QPA file instead.
Affected tests:
dEQP-VK.api.buffer_memory_requirements.*
Components: Vulkan
VK-GL-CTS issue: 3539
Change-Id: I6e2c9a04488ad86fcac83a27beedab6e0bf36cf3
Jeff Bolz [Tue, 1 Mar 2022 19:48:32 +0000 (13:48 -0600)]
Fix fragment_shader_interlock vs fragment_shading_rate tests
These tests were incorrectly running in a device created
with the NV extension enabled rather than the KHR.
Component: Vulkan
Affects: dEQP-VK.fragment_shader_interlock.*
Change-Id: Ic6005c0291dbd7c7bfd8befa135a7edcf253b91d
Graeme Leese [Fri, 22 Oct 2021 11:24:38 +0000 (12:24 +0100)]
Use standard type definitions on all platforms
These definitions are used in the Vulkan headers, so they are guaranteed
to be available on all platforms that the CTS supports. Trying to define
things in terms of other types is unnecessarily complicated. This is a
first small step towards using the standard types everywhere in the CTS.
Component: Framework
Affects: *
Change-Id: I492dff86e4f6521703feac1d6398d791a579b6fd
Graeme Leese [Wed, 14 Jul 2021 12:04:58 +0000 (13:04 +0100)]
Use common code for robustness2 tests
Where the code is the same for every stage, keep a single copy instead
of writing it out every time. This makes it easier to understand what is
different between the cases and why. It also means that we only have to
fix issues once instead of 6 times.
Components: Vulkan
Affects: dEQP-VK.robustness.robustness2.*
dEQP-VK.robustness.image_robustness.*
Change-Id: I8dcf6efee86165f78f657358a0f11fd6e2e2d163
Jeff Bolz [Mon, 28 Feb 2022 18:41:26 +0000 (12:41 -0600)]
Add tests to verify compressed texture downloads to non-zero mip levels
Component: Vulkan
Affects tests: dEQP-VK.texture.compressed*
Change-Id: I95544554787cc0d4d7da378da73315a0ed14c16c
Matthew Netsch [Thu, 10 Mar 2022 19:03:00 +0000 (19:03 +0000)]
Merge vk-gl-cts/vulkan-cts-1.3.1 into vk-gl-cts/main
Change-Id: I64f912202161fabd927dfc3d07db444b4ed0b5ef
Sławomir Cygan [Mon, 7 Mar 2022 19:32:16 +0000 (20:32 +0100)]
Add support for 'alternative' image view min lod calculation
According to ''Image Level(s) Selection" in spec, the image view
minLod value can be used as-is, or floor()-ed.
This change runs verification in affected test twice, once requiring
'preferred' method, falling back to 'alternative' if 'preferred' method fails.
Components: Vulkan
VK-GL-CTS Issue: 3556
Affects: dEQP-VK.texture.*min_lod*
Change-Id: I3af08485dbd99ad8d666c6f19aac3456130c375d
Thomas Spurden [Tue, 15 Feb 2022 11:12:00 +0000 (11:12 +0000)]
Update FBO tests to handle separate depth stencil in ES3
Extends FBO blit tests as they did not have cases for separate depth
stencil at all (as they are ES3 only).
Components: OpenGL
Affects: dEQP-GLES3.functional.fbo.*
Change-Id: Icc6dcd047db3b9c29baed0c5691766fb30242e31
Ricardo Garcia [Wed, 2 Mar 2022 16:04:07 +0000 (17:04 +0100)]
Avoid triangle fans in reinterpolation consistency tests
These causes problems for portability implementations and are not needed
for the test itself. This commit replaces them with triangle strips.
Affected tests:
dEQP-VK.pipeline.multisample_interpolation.reinterpolation_consistency.interpolate_at_centroid
dEQP-VK.pipeline.multisample_interpolation.reinterpolation_consistency.interpolate_at_sample
Components: Vulkan
VK-GL-CTS issue: 3545
Change-Id: I29aae73b8a74795b5297eb546a1d150a61e82c33
Matthew Netsch [Thu, 10 Mar 2022 18:39:28 +0000 (18:39 +0000)]
Merge vk-gl-cts/vulkan-cts-1.3.0 into vk-gl-cts/vulkan-cts-1.3.1
Change-Id: I14d7df956df08612d8b7827ee1fa9306a273af0b
Iván Briano [Tue, 8 Mar 2022 00:51:16 +0000 (16:51 -0800)]
Fix REQUIRE_FULL_SUBGROUPS_BIT tests with SPIR-V 1.6
When using Vulkan with SPIR-V 1.6 shaders, both FULL SUBGROUPS and ALLOW
VARYING flags are already implicitly set.
Components: Vulkan
VK-GL-CTS issue: 3541
Affects:
dEQP-VK.subgroups.size_control.compute.require_full_subgroups*spirv16
Change-Id: Ic8322292967101f0dbe739777625c2666a751fcb
Jason Ekstrand [Sat, 12 Feb 2022 18:24:45 +0000 (12:24 -0600)]
Use the correct resolve mode for dynamic rendering
For color attachments, you're required to use SAMPLE_ZERO for integer
and AVERAGE for non-integer.
Components: Vulkan
Affects: dEQP-VK.dynamic_rendering.suballocation.multisample_resolve.\*
Change-Id: Ia23c96f92d86283bc4094899f9473e2b35cf8490
Matthew Netsch [Thu, 10 Mar 2022 18:12:46 +0000 (18:12 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.8 into vk-gl-cts/vulkan-cts-1.3.0
Change-Id: I7a6e172719c8bd688e642a70e26dae42731715b1
Matthew Netsch [Thu, 10 Mar 2022 17:47:08 +0000 (17:47 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8
Change-Id: I2280d6ce320738b91994d4603600c38d5d1ba4ee
Matthew Netsch [Thu, 10 Mar 2022 17:20:41 +0000 (17:20 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7
Change-Id: If390df3efc02bff47f240c62e35d23e78c537b4c
Matthew Netsch [Thu, 10 Mar 2022 16:57:32 +0000 (16:57 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6
Change-Id: I31b3e7ec830aa10aa3c06ef61972a4e9c48dffc2
Matthew Netsch [Thu, 10 Mar 2022 16:34:26 +0000 (16:34 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5
Change-Id: I010f996f4b726f1f79e9bbe9b9f3e96131021f00
Boris Zanin [Tue, 1 Mar 2022 15:16:12 +0000 (16:16 +0100)]
Disable BC7 mode 8 pattern generation
Due to BC7 is underspecified for mode 8 (LSB=0x00) tests
should avoid testing it. This commit avoids generation
mode 8 BC7 compressed blocks.
Affects tests:
* dEQP-VK.*bc7*
Components: Framework, Vulkan
VK-GL-CTS issue: 3501
Change-Id: I222046f703fbead2bf5c64376defc6481989be24
Matthew Netsch [Fri, 4 Mar 2022 16:03:05 +0000 (16:03 +0000)]
Merge "Merge vk-gl-cts/vulkan-cts-1.3.1 into vk-gl-cts/main" into main
Samuel Iglesias Gonsálvez [Tue, 22 Feb 2022 11:20:26 +0000 (12:20 +0100)]
Add a check that drmFormatModifierTilingFeatures has at least one bit set
Components: Vulkan
VK-GL-CTS issue: 3532
Affected tests:
dEQP-VK.drm_format_modifiers.*
Change-Id: I7084208ffc5777b5d2bd5e16fe8e2bfb7df1adfa
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Ricardo Garcia [Tue, 22 Feb 2022 08:33:59 +0000 (09:33 +0100)]
Test primitive ID is not affected by culling
Check primitive culling does not affect the primitive ID.
New test:
dEQP-VK.rasterization.culling.primitive_id
Components: Vulkan
VK-GL-CTS issue: 3526
Change-Id: I348b981e775fefde33d128789324cd9ceb80e7a7
Antto Mäkinen [Fri, 7 Jan 2022 10:23:48 +0000 (12:23 +0200)]
Verify Depth/Stencil Write conditions
This commit adds new tests that verify the implementation properly
updates the depth and stencil buffers if a fragment is discarded.
VK-GL-CTS issue: 3345
New tests:
dEQP-VK.renderpass.depth_stencil_write_conditions.*
Components: Vulkan
Change-Id: Ifc65df616bf6d3febd89e494701f643b663fd9b2
Matthew Netsch [Thu, 3 Mar 2022 22:42:07 +0000 (22:42 +0000)]
Merge vk-gl-cts/opengl-cts-4.6.2 into vk-gl-cts/main
Change-Id: I4678511994ef6b055dc60d1fd78390b132e64a9b
Matthew Netsch [Thu, 17 Feb 2022 20:31:17 +0000 (12:31 -0800)]
Update KC CTS
Pull in fixes for !42, !46, and !47
Components: Framework
Change-Id: Ief4ea3aadef47d1b3e25a7fadd57cac8e8228d73
Matthew Netsch [Thu, 3 Mar 2022 22:40:56 +0000 (14:40 -0800)]
Merge vk-gl-cts/vulkan-cts-1.3.1 into vk-gl-cts/main
Change-Id: I7e8890118a65a9d1833a167cdbb80d87d2a38077
Charles Johnston [Fri, 25 Feb 2022 17:36:28 +0000 (12:36 -0500)]
Fix unsupported_features test.
The unsupported_features test will always fail when
EXT_pageable_device_local_memory is supported.
This change moves the code that disables pageableDeviceLocalMemory
into if(!enableAllFeatures) where it will function as expected.
Components: Vulkan
Affects: dEQP-VK.api.device_init.create_device_unsupported_features
VK-GL-CTS issue: 3543
Change-Id: If646473de3a3da743d706acbe5fcd2bd39973f1e
Sławomir Cygan [Fri, 25 Feb 2022 10:46:06 +0000 (11:46 +0100)]
Update glslang
Components: Vulkan
VK-GL-CTS Issue: 3536
Affects: dEQP-VK.*
Change-Id: I78e6dd5859b60d3e461e94dcdf32e760b2bc410a
Samuel Iglesias Gonsálvez [Tue, 22 Feb 2022 12:41:47 +0000 (13:41 +0100)]
Remove Cached as it is not needed on compute zero_initialize_workgroup_memory tests
Components: Vulkan
VK-GL-CTS issue: 3533
Affected tests:
dEQP-VK.compute.zero_initialize_workgroup_memory.*
Change-Id: Id966f7804cd25ac739c74f91298244dcfbc6a3e5
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Slawomir Cygan [Fri, 25 Feb 2022 15:55:50 +0000 (16:55 +0100)]
Use correct barriers in dEQP-VK.api.copy_and_blit.core.*no_cab
Use transfer read & transfer stage in barrier before using image as
source in vkCmdCopyImage & vkCmdCopyImage2
Components: Vulkan
Affects: dEQP-VK.api.copy_and_blit.*.resolve_image.*no_cab*
VK-GL-CTS Issue: 3542
Change-Id: I2c0d7971c8dcba1f1c01a902c271ce0556196fef
Matthew Netsch [Thu, 3 Mar 2022 19:53:25 +0000 (11:53 -0800)]
Merge vk-gl-cts/vulkan-cts-1.3.0 into vk-gl-cts/vulkan-cts-1.3.1
Change-Id: I13e284240b66cbaf79b74d2032bb6c5bf682ea5e
Antto Mäkinen [Tue, 15 Feb 2022 14:31:19 +0000 (16:31 +0200)]
Fix checkSupport in compressed texture sampling tests
Compressed texture sampling tests didn't properly check that images
using a compressed format can be created with the extended usage
flags. This commit fixes the check.
VK-GL-CTS issue: 3225
Affected Tests:
dEQP-VK.image.sample_texture.*
Components: Vulkan
Change-Id: I5a4af6ce3d9a351933403bceddb7765ea921901c
Slawomir Cygan [Wed, 23 Feb 2022 15:52:15 +0000 (16:52 +0100)]
Fix multiview dynamic rendering tests to enable dynamicRendering feature
The test uses own, custom VkDevice which is created without enabling
'dynamicRendering' feature - while this feature is used by the test.
The fix is to add VkPhysicalDeviceDynamicRenderingFeatures to the
chain of create infos used for getting and requesting features, so
the feature would be alwyas enabled if supported.
(The support of the feature is checked elsewhere).
Components: Vulkan
VK-GL-CTS Issue 3537
Affects: dEQP-VK.multiview.dynamic_rendering,*
Change-Id: Ibcc9a27e01c5deeff1ef90baf97eee5148244603
Matthew Netsch [Thu, 3 Mar 2022 18:26:20 +0000 (18:26 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.8 into vk-gl-cts/vulkan-cts-1.3.0
Change-Id: Ie2d3f521cea9ab73d6646abe1d2f9c66d2c6833e
Antto Mäkinen [Wed, 23 Feb 2022 12:18:08 +0000 (14:18 +0200)]
Fix checkSupport in compressed texture sampling tests
Compressed texture sampling tests didn't properly check that images
using a compressed format can be created with the extended usage
flags. This commit fixes the check.
VK-GL-CTS issue: 3225
Affected Tests:
dEQP-VK.image.sample_texture.*
Components: Vulkan
Change-Id: I1c077d29befac3d1b22c5ac2072518296f721f47
Matthew Netsch [Thu, 3 Mar 2022 18:00:13 +0000 (18:00 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8
Change-Id: Ib6194cc5034a2dada976ef494cf4c23fe997555f
Slawomir Cygan [Wed, 19 Jan 2022 11:51:01 +0000 (12:51 +0100)]
Don't adjust vertext buffer pointer in AS for indirect AS build.
Indirect build test programs offsets into the vertex buffer in the wr-asb
shader. This shader mimics vertex buffer pointer adjustment that is done in
AS build util.
Doing both adjustments (in shader and in AS) would cause out of bounds
accesses.
Authored-by: German, Leonid <leonid.german@intel.com>
Components: Vulkan
VK-GL-CTS Issue: 3540
Affects: dEQP-VK.ray_tracing_pipeline.indirect.build_structure
Change-Id: Iace6cf96d478641f412de3a3d0ec7ab001860f29
Ricardo Garcia [Thu, 17 Feb 2022 10:09:07 +0000 (11:09 +0100)]
Do not require full subgroups when requiring a work group size
Most subgroup tests that require a particular work group size use the
VK_PIPELINE_SHADER_STAGE_CREATE_REQUIRE_FULL_SUBGROUPS_BIT_EXT flag when
creating the shader stage, but that in turn forces the local size in the
X dimension to have certain values as required by the following VUID:
VUID-VkPipelineShaderStageCreateInfo-pNext-02757
There are specific tests to check full subgroups. The flag can be safely
omitted and these other tests should still work.
Affected tests:
dEQP-VK.subgroups.*compute*
Components: Vulkan
VK-GL-CTS issue: 3520
Change-Id: Ic20d07d3a875156f817066d040ddf5813aef311f
Matthew Netsch [Thu, 3 Mar 2022 17:34:09 +0000 (17:34 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7
Change-Id: I589df8409840b42f6c683dda57616b3ee8359128
Matthew Netsch [Thu, 3 Mar 2022 17:11:01 +0000 (17:11 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6
Change-Id: I3cf7433d510e0966f4f4d96971519aa849cecda0
Matthew Netsch [Thu, 3 Mar 2022 16:47:59 +0000 (16:47 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5
Change-Id: I8e72cdeaa8eed150a5548dd30871eccb09f31d5c
Max Andersson [Mon, 7 Feb 2022 11:31:13 +0000 (12:31 +0100)]
Mark gl_Position as invariant in invariance tests
Always marks gl_Position as invariant. The reason
for this is that there is no guarantee that it
will be computed in an invariant way otherwise,
which may cause test failures.
Affects:
dEQP-VK.glsl.invariance.*user_defined*
VK-GL-CTS issue: 3493
Change-Id: Ia172664b454182458dbd13146c2d1688ecf4161f
Thomas Spurden [Tue, 15 Feb 2022 11:12:00 +0000 (11:12 +0000)]
Update FBO tests to handle separate depth stencil in ES3
Extends FBO blit tests as they did not have cases for separate depth
stencil at all (as they are ES3 only).
Components: OpenGL
Affects: dEQP-GLES3.functional.fbo.*
Change-Id: Icc6dcd047db3b9c29baed0c5691766fb30242e31
Matthew Netsch [Thu, 24 Feb 2022 22:24:20 +0000 (22:24 +0000)]
Merge vk-gl-cts/vulkan-cts-1.3.1 into vk-gl-cts/main
Change-Id: I4f053e7f4bd1224d0666f72b5f4fd980bb6dd452
Matthew Netsch [Thu, 24 Feb 2022 22:01:36 +0000 (22:01 +0000)]
Merge vk-gl-cts/vulkan-cts-1.3.0 into vk-gl-cts/vulkan-cts-1.3.1
Change-Id: I71dd432beaf2983f9633000068f70a8ffba80bd0
Matthew Netsch [Thu, 24 Feb 2022 21:35:54 +0000 (21:35 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.8 into vk-gl-cts/vulkan-cts-1.3.0
Change-Id: I3195f9f15bfe716204d8908062f62528fd70f580
Matthew Netsch [Thu, 24 Feb 2022 21:09:39 +0000 (21:09 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8
Change-Id: I37050cfb18cd34bd869b26531623f3b881973a3b
James Fitzpatrick [Wed, 16 Feb 2022 15:23:21 +0000 (15:23 +0000)]
Increase tolerance in multiple interpolation tests
These tests test that if a shader has multiple varyings with different
interpolation modes each varying is interpolated using the correct
interpolation mode.
The test runs a shader with multiple varyings with different
interpolation modes and expects this to be an exact match with an image
produced by a similar shader with a single varying of the same
interpolation mode.
Due to floating point inaccuracies and different rounding, allow a 1
bit tolerance when comparing the images.
Affects: dEQP-VK.draw.*multiple_interpolation.*
Components: vulkan
VK-GL-CTS issue: 3522
Change-Id: Ib4a8ab527644de5e1c36ecd3e627afe5384cc6f6
Matthew Netsch [Thu, 24 Feb 2022 20:13:47 +0000 (12:13 -0800)]
Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7
Change-Id: I3e993e882a340347eea3582c1ba4feb3393b58fc
Matthew Netsch [Thu, 24 Feb 2022 19:07:22 +0000 (19:07 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6
Change-Id: I3ae2bc92916e5c4e2b525a085ba763d4c4b124d1
Ricardo Garcia [Fri, 11 Feb 2022 09:30:05 +0000 (10:30 +0100)]
Ray tracing: avoid skipping triangles and AABBs at the same time
This is illegal according to the spec.
Removed tests:
dEQP-VK.ray_tracing_pipeline.builtin.incomingrayflagsext.*pipelineskipboth*
dEQP-VK.ray_tracing_pipeline.builtin.incomingrayflagsext.*rayskiptriangles.pipelineskipaabbs*
dEQP-VK.ray_tracing_pipeline.builtin.incomingrayflagsext.*rayskipaabbs.pipelineskiptriangles*
dEQP-VK.ray_tracing_pipeline.builtin.incomingrayflagsext.*rayskipboth*
Components: Vulkan
VK-GL-CTS issue: 3508
Change-Id: Iebc50b4b546abc36b2607eeb7c5cbb900de35fcb
Ari Suonpaa [Thu, 10 Feb 2022 05:31:05 +0000 (07:31 +0200)]
Test message passing using permuted indices
This change also adds support for a custom checkSupport
callback when creating Amber test cases.
VK-GL-CTS Issue: 3390
New tests:
dEQP-VK.memory_model.message_passing.permuted_index.*
Components: Vulkan, Framework
Change-Id: I68c13e0106acc4c2a2f7541c4acdf3274009ad7b
michal_jakubek [Tue, 1 Feb 2022 09:57:21 +0000 (10:57 +0100)]
Reduce and amortize memory allocations in RT utilities
Affects:
dEQP-VK.ray_tracing*
dEQP-VK.ray_query*
Components: Framework
VK-GL-CTS issue: 3487
Change-Id: Ia42a5745ea302c27027d13b95d0ae8d565feed1a
Mike Blumenkrantz [Wed, 23 Feb 2022 22:14:31 +0000 (17:14 -0500)]
fix GL46 vector construction sizing
the vector here is 4 rows of 2 components, but std::vector.size()
returns the total vector length (8), so divide by 2 to get the correct
length here and avoid out-of-bounds reads
Affects:
KHR-GL46.shaders30.glsl_constructors*
Components: OpenGL
VK-GL-CTS Issue: 3538
Change-Id: I8a07957722bf8b49fdcee14fe109d284f8cb9e27
Graeme Leese [Mon, 21 Feb 2022 16:59:15 +0000 (16:59 +0000)]
Fix arguments for image test reference generation
The default arguments were incorrect in fd819aa, causing some versions
of the generateReferenceImage function to fail.
There were overloads with 2 and 3 arguments, one was added so 3 or 4
arguments were valid, but only the 3 argument version allowed a default.
Thus the 2 argument versions carried on working, but the 3 argument
calls now matched the old 2 argument version with the new parameter,
which mangled the type into a bool and was generally broken. Here the
default is added to the other prototype as well, causing that to be the
preferred overload and everything goes back to how it was before.
Components: Vulkan
VK-GL-CTS issue: 3372
Affects: dEQP-VK.image.format_reinterpret.*
dEQP-VK.image.* (possibly)
Change-Id: I1d6da992570988018ebff407697c8ef9a62c823f
Matthew Netsch [Thu, 17 Feb 2022 22:54:23 +0000 (22:54 +0000)]
Merge vk-gl-cts/github-main into vk-gl-cts/main
Change-Id: I8fb9385f422ee19c98b98d1c546e37642afd2bf6
Dave Evans [Wed, 8 Sep 2021 17:03:23 +0000 (12:03 -0500)]
Fix waived tests being able to crash suite
- No longer calls TestCase constructor for waived tests
Components: Vulkan
VK-GL-CTS Issue: 3478
Change-Id: Iacc9049b71c8b479cc2c3e1e5b2f0cbea4c129af
(cherry picked from commit
44da8ba520e4929547a8b1daf90cf43fe42616fd)
Matthew Netsch [Thu, 17 Feb 2022 22:17:26 +0000 (22:17 +0000)]
Merge vk-gl-cts/vulkan-cts-1.3.1 into vk-gl-cts/main
Change-Id: Ib11bcbeddd3910d25918cd9404c581d129cf5899
Piers Daniell [Thu, 10 Feb 2022 19:28:26 +0000 (12:28 -0700)]
Fix OOB memory copy in dEQP-VK.image.sample_cubemap.* tests
The copy was reading beyond the end of the vector.
Affects:
dEQP-VK.image.sample_cubemap.write_face_0
Components: Vulkan
VK-GL-CTS issue: 3515
Change-Id: I6629968ce96b25f413d33be99ae693ce69b77421
Samuel Iglesias Gonsálvez [Tue, 8 Feb 2022 13:58:10 +0000 (14:58 +0100)]
Only pass VkImageViewMinLodCreateInfoEXT when minLod is != 0.0f
We avoid to passing VkImageViewMinLodCreateInfoEXT when creating
the image view when the tests are actually not testing this.
Therefore, the drivers that don't support this extension won't
see errors when processing VkImageViewCreateInfo.
Components: Vulkan
VK-GL-CTS issue: 3094
Affected tests:
dEQP-VK.texture.mipmap.*
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Change-Id: I09110ef2e6ebfb6f55c084eda85359ebf5cfa742
Ricardo Garcia [Tue, 8 Feb 2022 15:20:26 +0000 (16:20 +0100)]
Fix uniform subgroup size test
The uniform subgroup size test was requiring the subgroup size to be
uniform across several vkCmdDispatch calls, which is not required by the
spec. Furthermore, both dispatch calls were using different local sizes
in the shader.
In addition, these changes add a missing barrier from compute to host.
Components: Vulkan
VK-GL-CTS issue: 3505
Affected tests:
dEQP-VK.subgroups.multiple_dispatches.uniform_subgroup_size
Change-Id: Icea2479da1793f0b24f6fb8a4dd60d477ef8a775
Piers Daniell [Mon, 14 Feb 2022 16:48:32 +0000 (09:48 -0700)]
Remove Roadmap 2022 requirement
The descriptorBindingUniformBufferUpdateAfterBind feature was
accidently required as part of the Roadmap 2022 in the Vulkan 1.3
release spec. It was subsequently fixed in the spec by
https://gitlab.khronos.org/vulkan/vulkan/-/merge_requests/5051
so also need to be removed for the next CTS release.
Affects:
dEQP-VK.api.info.profiles.roadmap_2022
Components: Vulkan
VK-GL-CTS issue: 3519
Change-Id: Id01bab7e5c48945d28b1a7fddeeecfdf7bdc36b2
Alejandro Piñeiro [Wed, 9 Feb 2022 10:54:53 +0000 (11:54 +0100)]
TimeLine semaphores tests should check for the feature
Recently new timeleine semaphores were added, but they were not
checking if time line semaphores are supported.
Affects:
dEQP-VK.api.external.semaphore.opaque_fd.info_timeline
dEQP-VK.api.external.semaphore.sync_fd.info_timeline
Components: Vulkan
VK-GL-CTS issue: 3510
Change-Id: Ic1029d82df6353fe831f0c791e08b7f4ca365b62
Ricardo Garcia [Wed, 16 Feb 2022 08:56:47 +0000 (09:56 +0100)]
Test more depth/stencil formats in load_store_op_none tests
Some load_store_op_none tests check some aspects of those operations
work with depth/stencil formats, but they skip formats which only have
depth or stencil aspects. This could be interesting, specially related
to dynamic rendering. This change adds tests that use other formats,
including:
- D16_UNORM, which has mandatory support in the spec.
- S8_UINT, which is the only format having only the stencil aspect.
New tests:
dEQP-VK.*.load_store_op_none.depth_d16_unorm*
dEQP-VK.*.load_store_op_none.depth_d32_sfloat*
dEQP-VK.*.load_store_op_none.stencil_s8_uint*
Components: Vulkan
VK-GL-CTS issue: 3466
Change-Id: I7ce4e8b82a98eb07ffec29b63ccad0fb93759845
Ricardo Garcia [Wed, 16 Feb 2022 08:51:52 +0000 (09:51 +0100)]
Check dynamic viewports with depth clip control
When the view volume depth range changes from [0,1] to [-1,1] there may
be implementations that need to be more careful when emitting
instructions that also take the viewport into account. As the viewport
can be made dynamic despite using depth clip control we need to check
that combination of factors.
Components: Vulkan
VK-GL-CTS issue: 3480
New tests:
dEQP-VK.pipeline.depth.depth_clip_control.*viewport*
Affected tests:
dEQP-VK.pipeline.depth.depth_clip_control.*
Change-Id: I5cf5b7998fd77a0569317f32e480f5a07d5ab969
Juha Heiskanen [Tue, 8 Feb 2022 13:21:36 +0000 (15:21 +0200)]
Add tests for zero initialize workgroup memory
New tests ensure that the workgroup initialization happens before
workgroup entry point.
Components: Vulkan
VK-GL-CTS Issue: 3326
New tests: dEQP-VK.compute.zero_initialize_workgroup_memory.shared_memory_blocks.*
Change-Id: I7ed42cb82a8ac6b36eeec3471369a31dad6c3e89
Matthew Netsch [Thu, 17 Feb 2022 21:40:39 +0000 (21:40 +0000)]
Merge vk-gl-cts/opengl-cts-4.6.2 into vk-gl-cts/main
Change-Id: Id30d32f99b11b181ad8e81dc261d47d8b7431ddd
Matthew Netsch [Thu, 17 Feb 2022 21:13:30 +0000 (21:13 +0000)]
Merge vk-gl-cts/opengl-cts-4.6.1 into vk-gl-cts/opengl-cts-4.6.2
Change-Id: I8b8cddd34f78dd0b134ae732b3c539cf503be087
Matthew Netsch [Thu, 17 Feb 2022 20:34:23 +0000 (20:34 +0000)]
Merge vk-gl-cts/opengl-cts-4.6.0 into vk-gl-cts/opengl-cts-4.6.1
Change-Id: I04ec68c36522b8082694e1f3f849ffbdafd34ea7
Mike Blumenkrantz [Tue, 18 Jan 2022 17:48:01 +0000 (12:48 -0500)]
fix invalid buffer bind size in KHR-GL46.sparse_buffer_tests.BufferStorageTest
m_sparse_bo_data_size is the size of the sparse data, but this may exceed
the value returned by GL_MAX_UNIFORM_BLOCK_SIZE, which (should) cause
the test to fail on any driver where the sparse data size is larger than
the max uniform block size
instead, use the size of the data being passed to the shaders directly
so that this can never be too large
Affects:
KHR-GL46.sparse_buffer_tests.BufferStorageTest
Components: OpenGL
VK-GL-CTS Issue: 3468
Change-Id: Ie9d3de5760773e9b1b3b24e8af77976e7a54036c
Matthew Netsch [Thu, 17 Feb 2022 20:27:22 +0000 (12:27 -0800)]
Merge vk-gl-cts/vulkan-cts-1.3.0 into vk-gl-cts/vulkan-cts-1.3.1
Change-Id: Ib56927d23f4d53fd739cd57dde59766e2dd12f5c
James Fitzpatrick [Sun, 6 Feb 2022 10:33:53 +0000 (10:33 +0000)]
Update mandatory features for Vulkan1.3
Vulkan 1.3 only requires
descriptorBindingInlineUniformBlockUpdateAfterBind to be supported if
descriptor indexing is supported.
Vulkan 1.3 also made a number of Vulkan 1.2 features mandatory, so add
these to the mandatory feature check for 1.3 implementations.
Affects:
dEQP-VK.api.info.vulkan1p3.feature_extensions_consistency
dEQP-VK.info.device_mandatory_features
Compontents: vulkan
VK-GL-CTS issue: 3496
Change-Id: I79a6df710b7770171b7a1128cdec30fa30d6a456
Matthew Netsch [Thu, 17 Feb 2022 19:48:55 +0000 (19:48 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.8 into vk-gl-cts/vulkan-cts-1.3.0
Change-Id: I67ca22b6f7d5e71d139482417f8ac5f30f6a1f81
Connor Abbott [Fri, 21 Jan 2022 17:28:59 +0000 (18:28 +0100)]
Always use workaround for blitting to/from BC6H
Qualcomm's implementation of cubic filtering can give large inaccuracies
when texel values within the 4x4 grid are very big and very small,
especially if the values around the "edge" are much bigger then the
values at the center. This can happen with randomly-generated blocks for
floating-point compressed formats like BC6H. There was already a
workaround for similar issues seen blitting to e5b9g9r9, extend it to
all formats to fix this.
Components: Vulkan
VK-GL-CTS Issue: 3450
Affects:
dEQP-VK.api.copy_and_blit.core.blit_image.all_formats.color.2d.bc6h_ufloat_block.*
dEQP-VK.api.copy_and_blit.core.blit_image.all_formats.color.2d.*.bc6h_ufloat_block.*
Change-Id: I3445ecba2ebc9e4e52e26111a109eaf79f611bd8
Will Lueking [Wed, 9 Feb 2022 20:34:10 +0000 (12:34 -0800)]
Fix build error on gcc 4.8
Changes a copy initializtion to a direct initialization.
GCC 4.8 complains about the copy initialization of
const tcu::StringTemplate meshTemplate = meshTemplateStr.str();
but not the direct initialization of
const tcu::StringTemplate taskDataDecl(taskDataDeclTemplate);
Replacing the copy initialization with the direct initialization fixes
the build issue.
Components: Vulkan
Affects: dEQP-VK.mesh_shader.nv.*
VK-GL-CTS issue: 3511
Change-Id: I9de57bc28e56da8288894e32c938fea0cd165927
Ricardo Garcia [Wed, 9 Feb 2022 10:09:33 +0000 (11:09 +0100)]
Fix load_store_op_none dynamic rendering validation error
The test case below was using a multisample color attachment with
integer format and specified VK_RESOLVE_MODE_AVERAGE_BIT as the resolve
mode, which is illegal. The purpose of the test is different, and the
resolve mode is not relevant, so it can be switched to one of the legal
values.
Affected tests:
dEQP-VK.dynamic_rendering.suballocation.load_store_op_none.color_load_op_none_store_op_none_resolve
Components: Vulkan
VK-GL-CTS issue: 3509
Change-Id: I42cf959f67d80694aee37c19c966ab037609d44f
Lionel Landwerlin [Sun, 30 Jan 2022 22:29:43 +0000 (00:29 +0200)]
Fix dynamic rendering multisampling test layout transition operation
Those tests are failing on Intel/Mesa because they're not
transitioning the right mip level from COLOR_ATTACHMENT to TRANFER_DST
layouts.
Component: Vulkan
Affects: dEQP-VK.dynamic_rendering.*.multisample_resolve.*resolve_level*
VK-GL-CTS issue: 3482
Change-Id: I6dcdf6a9a72d96c559acd28f0114b308267d3cb8
Jason Ekstrand [Fri, 11 Feb 2022 00:20:08 +0000 (18:20 -0600)]
Use the correct stencilInitialLayout for unused resolve targets with separate stencil
The initial stencil layout was always being specified as UNDEFINED even
though, in the unused resolve target case, we clear it with
vkCmdClearDepthStencilImage() and need an initial layout of
TRANSFER_SRC_OPTIMAL.
Components: Vulkan
Affects: dEQP-VK.renderpass2.depth_stencil_resolve.*separate_layouts.*
Change-Id: Id0e0dcd2066d02758875a786c5634b23c893f7a2
Matthew Netsch [Thu, 17 Feb 2022 19:22:09 +0000 (19:22 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8
Change-Id: If3af09eda7b56dbfa484de49f06401ff53fda906
Ricardo Garcia [Thu, 10 Feb 2022 11:22:55 +0000 (12:22 +0100)]
Avoid invalid reportIntersectionEXT hit kinds in subroup tests
Some subgroup ray tracing tests were using invalid values when calling the
reportIntersectionEXT function.
Affected tests:
dEQP-VK.subgroups.*ray_tracing*
Components: Vulkan
VK-GL-CTS issue: 3513
Change-Id: I96310c32e1e5b4fde157570f0a3131d61f5a125c
Matthew Netsch [Thu, 17 Feb 2022 18:55:27 +0000 (18:55 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7
Change-Id: Iba2783e1e1e2bcb839c075dcf4d2b1dbcfbdf43f
Mika Väinölä [Thu, 3 Feb 2022 08:15:19 +0000 (10:15 +0200)]
Replace bitwise or with logical
This fixes a Clang 14 bitwise-instead-of-logical warning.
Affects:
dEQP-VK.robustness.robustness2.*
dEQP-VK.robustness.image_robustness.*
Components: Vulkan
VK-GL-CTS issue: 3492
Change-Id: I803ed494ed218b404a2285ed891d41170d66abfe
Matthew Netsch [Thu, 17 Feb 2022 18:32:10 +0000 (18:32 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6
Change-Id: I4adbb2e30086c26c8ad3312a6086e267c2840d42
Ricardo Garcia [Thu, 10 Feb 2022 10:49:21 +0000 (11:49 +0100)]
Avoid invalid reportIntersectionEXT hit kinds in mem guarantee tests
Only values in the [0, 127] range are allowed with reportIntersectionEXT.
Affected tests:
dEQP-VK.ray_tracing_pipeline.memguarantee.*
Components: Vulkan
VK-GL-CTS issue: 3513
Change-Id: I54669b5b7e2c0f696bbbf3e5a477b1533e4f4488
Matthew Netsch [Thu, 17 Feb 2022 18:08:53 +0000 (18:08 +0000)]
Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5
Change-Id: Ib8646342ae0c6c9df3c7f2fa78c24643c10d5e96
Ilkka Saarelainen [Wed, 2 Feb 2022 13:37:18 +0000 (15:37 +0200)]
Fix validation errors in compute tests
This CL fixes various validation errors in compute tests.
Affects:
dEQP-VK.compute.basic.image_barrier*
dEQP-VK.compute.cooperative_matrix.*
Components: Vulkan
VK-GL-CTS Issue: 3451
Change-Id: I33037a50d10a5cf8098c33b55ba42900f97f0019
Matthew Netsch [Tue, 15 Feb 2022 16:51:37 +0000 (11:51 -0500)]
Merge pull request #305 from akihikodaki/do-not-directly-use-openglcts
Do not directly use openglcts
Akihiko Odaki [Tue, 15 Feb 2022 08:17:52 +0000 (17:17 +0900)]
Do not directly use openglcts
Directly specifying openglcts in CMake results in the following error:
> CMake Error at modules/common/subgroups/CMakeLists.txt:42 (PCH):
> Unknown CMake command "PCH".
Matthew Netsch [Fri, 11 Feb 2022 21:36:23 +0000 (21:36 +0000)]
Merge vk-gl-cts/github-main into vk-gl-cts/main
Change-Id: I69819f435b185677aa7817fae9284b1d9c3e95a8
Matthew Netsch [Thu, 10 Feb 2022 20:07:17 +0000 (20:07 +0000)]
Merge vk-gl-cts/vulkan-cts-1.3.1 into vk-gl-cts/master
Change-Id: Ibee17f19d6dca7cf6de24e54363ded04e367c1ea
Alejandro Piñeiro [Tue, 8 Feb 2022 10:56:32 +0000 (11:56 +0100)]
Increase error threshold for blitting non-nearest compressed formats
The error threshold for linear filtering with compressed formats is
too low, making some ASTC tests failingn with some drivers.
Affects:
dEQP-VK.api.copy_and_blit.core.blit_image.all_formats.color.2d.*.*.general_general_linear
Components: Vulkan
VK-GL-CTS issue: 3504
Change-Id: I917c3e5e6a0af26c59b004ac610378f4842cf1f7
Juha Heiskanen [Wed, 19 Jan 2022 10:26:28 +0000 (12:26 +0200)]
Resolve multisample image tests with renderArea < attachment size
Multisample resolve tests where a render area is less than the
attachment size.
Components: Vulkan
VK-GL-CTS Issue: 3293
New tests: dEQP-VK.pipeline.multisample.resolve.renderpass_renderarea.*
Change-Id: I908b75094e7397b6e932298a736c3ef91ee5c1be
Samuel Iglesias Gonsálvez [Tue, 14 Dec 2021 13:31:49 +0000 (14:31 +0100)]
Added tests to check vkGetPhysicalDeviceImageFormatProperties with VK_IMAGE_CREATE_EXTENDED_USAGE_BIT
Components: Vulkan
VK-GL-CTS issue: 3414
Added tests:
dEQP-VK.image.extended_usage_bit_compatibility.*
Change-Id: I392b785d97d4468245b8fad860656418005f785e
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Matthew Netsch [Thu, 10 Feb 2022 19:53:51 +0000 (14:53 -0500)]
Merge pull request #283 from imirkin/master
Require unsized half-float textures to be renderable
Ilkka Saarelainen [Tue, 18 Jan 2022 11:01:49 +0000 (13:01 +0200)]
Remove duplicate makeComputePipeline definitions
This CL removes the duplicate makeComputePipeline definitions.
Only one definition is left in vkObjUtil.hpp/cpp.
Components: Vulkan, Framework
Affects:
dEQP-VK.*
VK-GL-CTS Issue: 3316
Change-Id: I52f316e8c1a1cc00c788979e5281b1f45be3ad81
Antto Mäkinen [Wed, 5 Jan 2022 14:42:10 +0000 (16:42 +0200)]
Add a batch of GraphicsFuzz coverage tests
This commit adds a batch of GraphicsFuzz coverage tests.
Components: Vulkan
New Tests:
dEQP-VK.graphicsfuzz.cov-bitwise-shift-right-full-bits-no-effect-clamp
dEQP-VK.graphicsfuzz.cov-conditions-empty-blocks-index-array-one-divided-by-findlsb
dEQP-VK.graphicsfuzz.cov-discard-condition-loop-same-condition-again
dEQP-VK.graphicsfuzz.cov-global-loop-counter-for-loop-function-call-inside-never-called
dEQP-VK.graphicsfuzz.cov-ivec-from-uniform-float-shift-right-add-components
dEQP-VK.graphicsfuzz.cov-loop-increment-integer-set-output-color-break
dEQP-VK.graphicsfuzz.cov-multiple-loops-same-condition-always-false-global-loop-counter
dEQP-VK.graphicsfuzz.cov-nested-loops-increase-integer-dot-product
Change-Id: I8be6b8c3d5ee2a3b2803b51de4f5e38591528935