James Fitzpatrick [Wed, 3 Nov 2021 14:31:50 +0000 (14:31 +0000)]
Fix invocation count verification
The shaders in these tests track the number of invocations executed and
verify that against the expected number of invocations for the test.
The spec only guarentees that each fragment is executed at least once.
So an implementation is free to execute each fragment multiple times
if required (for example, an implementation might required a shader to
be run at sample rate).
Change this verification to only check against the minmum number of
invocations.
Affects: dEQP-VK.fragment_operations.early_fragment.samplemask_*
Components: Vulkan
VK-GL-CTS issue: 3278
Change-Id: I93f1aa42dd6a13486bed044e361671e5a2f2cb07
Iván Briano [Mon, 18 Oct 2021 20:32:00 +0000 (13:32 -0700)]
Slightly increase tolerance for fp64 determinant test
VK-GL-CTS issue: 3229
Components: Vulkan
Affects:
dEQP-VK.glsl.builtin.precision_double.determinant.compute.mat3
Change-Id: I2e4d239a33e4c82e328a3df90e7a6c38056db4d8
(cherry picked from commit
8c05f50808568561b962583e589c6781cbfccd67)
Matthew Netsch [Fri, 5 Nov 2021 14:51:30 +0000 (10:51 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.3.0
Change-Id: Ia059b652611b41999709074e2c9e3830ef6fac1f
Lionel Landwerlin [Wed, 6 Oct 2021 14:45:53 +0000 (17:45 +0300)]
Prevent RT tests from going over VkPhysicalDeviceVulkan11Properties::maxMemoryAllocationSize
VK-GL-CTS issue: 3064
Component: Vulkan
Affects:
dEQP-VK.ray_tracing_pipeline.misc.max*
Change-Id: Ia42f5530206eeeb94fa634ad84a213c2db558946
Mika Väinölä [Thu, 21 Oct 2021 07:46:33 +0000 (10:46 +0300)]
Pause XFB during pipeline switch
The per pipeline provoking vertex tests call vkCmdBindPipeline while
transform feedback is active which is not allowed.
Affects:
dEQP-VK.rasterization.provoking_vertex.*
Components: Vulkan
VK-GL-CTS issue: 3207
Change-Id: I30edf010179ff9629fc90053f14d228d828551a2
Matthew Netsch [Fri, 5 Nov 2021 13:59:25 +0000 (09:59 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7
Change-Id: I81ba38a964a81a6266d2a6286cf60a42ed027885
Piotr Byszewski [Tue, 2 Nov 2021 11:13:32 +0000 (12:13 +0100)]
Portability: handle robustBufferAccess
When the VK_KHR_portability_subset extension is advertised,
robustBufferAccess is no longer a required feature.
Components: Vulkan
VK-GL-CTS issue: 3267
Affets:
dEQP-VK.robustness.*
Change-Id: I957da2a36c40662c99c0b01774a8bf03ef3f3890
Matthew Netsch [Thu, 4 Nov 2021 20:45:41 +0000 (16:45 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6
Change-Id: I53d429b0dfc4a62a33949748ef4ce0a804f71814
Matthew Netsch [Thu, 4 Nov 2021 20:45:39 +0000 (16:45 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5
Change-Id: I1c61bd275a7161d071ad79aa493c6634582391a2
Matthew Netsch [Thu, 4 Nov 2021 20:45:32 +0000 (16:45 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.3 into vk-gl-cts/vulkan-cts-1.2.4
Change-Id: Ice8e3ca5f949d6c134f91a7d5317ab5892d5147d
Matthew Netsch [Mon, 1 Nov 2021 14:56:44 +0000 (10:56 -0400)]
Allow Vulkan CTS 1.2.7.2
Affects: dEQP-VK.api.driver_properties.properties
Components: Vulkan
Change-Id: I6075a82eab8f347ef5b9523a10d6e6621163bfbc
Jason Ekstrand [Mon, 25 Oct 2021 22:25:16 +0000 (17:25 -0500)]
Fix layout transitions in dynamic_rendering tests
With dynamic rendering, we no longer have initial and final layout
transitions provided automatically as part of vkCmdBeginRendering() and
vkCmdEndRendering(). Instead, it's the responsibility of the client to
get the images into the layout passed to vkCmdBeginRendering(). When
translating from Vulkan 1.0 style subpasses, this means we need to do a
pipeline barrier before and after rendering.
Components: Vulkan
VK-GL-CTS Issue: 3251
Affects: dEQP-VK.renderpass_with_dynamic_rendering.*
Change-Id: I56214198a8b3d81e3030164de045b8ca03a44f88
Piers Daniell [Thu, 28 Oct 2021 17:35:54 +0000 (11:35 -0600)]
Fix for Vulkan 1.3 implementations that don't expose VK_EXT_ycbcr_2plane_444_formats
The VK_EXT_ycbcr_2plane_444_formats extension was only partially promoted
to Vulkan 1.3. In this case the features structure was not promoted
and is not required for Vulkan 1.3 implementations to support it.
This CL modifies the generated framework so that the
VkPhysicalDeviceYcbcr2Plane444FormatsFeaturesEXT is not required to
be supported.
Affects:
dEQP-VK.api.info.get_physical_device_properties2.features
dEQP-VK.*2plane_444*
Components: Vulkan
VK-GL-CTS issue: 3252
Change-Id: I22d95869febd5f8a8c9900e8a659f1609f4c2264
Piers Daniell [Thu, 28 Oct 2021 16:46:40 +0000 (10:46 -0600)]
Skip negative vkCreateBuffer(size_max_uint64) test for maint4
With VK_KHR_maintenance4 it become invalid usage to call vkCreateBuffer
with a size larger than VkPhysicalDeviceMaintenance4Properties.maxBufferSize.
The dEQP-VK.api.buffer.basic.size_max_uint64 did just this even on maint4
implementation, which is invalid usage.
This CL marks this test as unsupported for implementations that support
VK_KHR_maintenance4.
Affects:
dEQP-VK.api.buffer.basic.*
Components: Vulkan
VK-GL-CTS issue: 3238
Change-Id: I4754c3af4ef253e50226a31a466bfe5772e65652
Matthew Netsch [Fri, 29 Oct 2021 18:00:47 +0000 (14:00 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.3.0
Change-Id: Ib88986ca0e77d8521e67413abeac510d7fe6cb2a
Paul Thomson [Mon, 25 Oct 2021 18:41:00 +0000 (19:41 +0100)]
Update SPIRV-Headers
The SPIRV-Headers version for vulkan-cts-1.2.6 was ahead of 1.2.7. This
change brings SPIRV-Headers in 1.2.7 up to the same version.
Components: Vulkan
Change-Id: Ifff93c5251183d3a6ee3ef3f508e88274332116f
Matthew Netsch [Fri, 29 Oct 2021 14:41:15 +0000 (10:41 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7
Change-Id: I7535fbad6813c7a6dfec1324a2656329de792ebe
Mohankumar Nekkarakalaya [Tue, 12 Oct 2021 18:58:22 +0000 (11:58 -0700)]
Remove 64-bit int usage in dynamic_indexing test shaders
- "OpCapability Int64" is not a mandatory extension
- Replace with v2i32
Components: Vulkan
VK-GL-CTS Issue: 3205
Affected tests:
dEQP-VK.ray_tracing_pipeline.acceleration_structures.dynamic_indexing.dynamic_indexing
dEQP-VK.ray_query.acceleration_structures.dynamic_indexing.dynamic_indexing
Change-Id: I04b7cf6309b42ef4572e5c79c8e9233aee2edec7
Matthew Netsch [Fri, 29 Oct 2021 13:52:23 +0000 (09:52 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6
Change-Id: I8a9f12e6adc0daf962a4f6f05d28216db4880ee8
Ricardo Garcia [Thu, 14 Oct 2021 11:27:46 +0000 (13:27 +0200)]
Check shaderInt64 in workgroup memory explicit layout alias tests
A few workgroup memory explicit layout alias tests were not checking
shaderInt64 while using 64-bit integers in shaders.
Affected tests:
dEQP-VK.compute.workgroup_memory_explicit_layout.alias.*u64*array*
dEQP-VK.compute.workgroup_memory_explicit_layout.alias.*array*u64*
Components: Vulkan
VK-GL-CTS issue: 3131
Change-Id: I1911018c471e9548603d5de987625163bb356eb9
Mohankumar Nekkarakalaya [Thu, 21 Oct 2021 00:14:08 +0000 (17:14 -0700)]
Allow NaN result when result exceeds limit
Based on spec if the result is too large to be represented
in the floating point the result is undefined.
Issue was found with ldexp. Since fix is applicable in general
updated the isFinite() to take maxValue and checks the interval value
with the allowed max value in the context (and similarly for
negative numbers). If the result is beyond the 32-bit
floating-point representation expected result is unbounded
Components: Vulkan, OpenGL, Framework
VK-GL-CTS Issue: 3138
Affected tests:
dEQP-VK.glsl.builtin.function.*
dEQP-VK.glsl.builtin.precision.*
dEQP-GLES31.functional.shaders.builtin_functions.precision.*
dEQP-GLES31.functional.shaders.builtin_functions.common.fma.*
Change-Id: Iba27d6b0d7d0bc433a1d0c055cef6f6a61b239ba
Graeme Leese [Fri, 27 Aug 2021 12:46:19 +0000 (13:46 +0100)]
Accept NaN as valid in atan2 precision tests
The Vulkan CTS has been updated
(https://gerrit.khronos.org/c/vk-gl-cts/+/7857) so that it is not
required to generate correct results for atan(inf, inf), and this makes
the corresponding change in GL. The GLSL spec is not as clear on what is
required here, but does say:
Built-in functions not listed above and not defined as equations of
the above have undefined precision. These include, for example, the
trigonometric functions
In general, I think that it is expected that the default state of Vulkan
is a match for GL.
Components: OpenGL
VK-GL-CTS issue: 3073
Affects: dEQP-GLES3.functional.shaders.builtin_functions.precision.atan2.highp_*
Change-Id: Ifae97a2e86962f5aff68d5eed09ba38069445468
Matthew Netsch [Thu, 28 Oct 2021 21:57:06 +0000 (17:57 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5
Change-Id: I54315f34095f396294bab85b0da4a1e3cf38aac8
Matthew Netsch [Thu, 28 Oct 2021 21:57:00 +0000 (17:57 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.3 into vk-gl-cts/vulkan-cts-1.2.4
Change-Id: I2f2be91bd696c7de227f804202f9a8eb222cab03
Ricardo Garcia [Fri, 8 Oct 2021 07:50:00 +0000 (09:50 +0200)]
Fix VK_KHR_shader_clock feature checks
The shader clock tests were not checking if the implementation supported
64-bit integers for tests that required the feature and they were
checking supported features at test instance creation time instead of
the specific checkSupport method for test cases.
Affected tests:
dEQP-VK.glsl.shader_clock.*
Components: Vulkan
VK-GL-CTS issue: 3187
Change-Id: I0dc86a4fae619c0d27628d97f0ac11c5def5ff4f
Juha Heiskanen [Tue, 19 Oct 2021 11:38:09 +0000 (14:38 +0300)]
Fix dEQP-VK.tessellation.geometry_interaction.limits.output_required_max_geometry validation error
Adjust numSliceNodesComponentLimit in order to keep
m_numGeometryPrimitivesPerInvocation within minimum limits defined in
the spec.
Components: Vulkan
VK-GL-CTS Issue: 3180
Affects:
dEQP-VK.tessellation.geometry_interaction.limits.output_required_max_geometry
Change-Id: I72ca43bec27256cb75a11d75aab91b9b565be9ab
Jari Komppa [Tue, 5 Oct 2021 09:52:03 +0000 (12:52 +0300)]
Move log prints from test-ctor to test-inst-ctor
Some of the tesselation tests were printing values into the output log
from the test constructor, which meant that said prints end up in more
or less random place in the log. Moved those prints to the test instance
instead, so the information now resides within the test output.
The prints were also done in the amber-verify run mode.
Affects:
dEQP-VK.tessellation.geometry_interaction.scatter.*
dEQP-VK.tessellation.geometry_interaction.limits.*
Components: Vulkan
VK-GL-CTS issue: 3057
Change-Id: Ic5b2ce59369d1c20f25b8dc1c6cb594baac74f8f
Sławomir Cygan [Thu, 21 Oct 2021 13:18:19 +0000 (15:18 +0200)]
Fix stencil rendering in draw_with_dynamic_rendering.multi_draw tests
This fixes few issues with handling depth/stencil attachment in this test:
Stencil attachment is now passed to rendering info, alongside depth attachment.
A barrier transitioning depth stencil subresource from UNFEFINED layout is added.
A barrier initializing color subresource is fixed to be executed with desition stage
that involves color.
Components: Vulkan
VK-GL-CTS Issue: 3235
Affects: dEQP-VK.draw_with_dynamic_rendering.multi_draw.*
Change-Id: I2b1a0bb04dae24328ab798a9e436e1000db901de
jaxl [Wed, 15 Sep 2021 03:14:19 +0000 (11:14 +0800)]
Add missing barrier to dynamic rendering tests
Vulkan spec requires driver to do automatic layout transition when
subpass of render pass begins. However, for dynamic rendering,
there is no such requirement in spec, so the application should
transfer image layout.
Affected tests:
VK.renderpass_with_dynamic_rendering.suballocation.multisample_resolve.*
Components: Vulkan
VK-GL-CTS issue: 3162
Change-Id: Id3bb8fa83ac4f891150f1a8f74bf61e7d7d5e854
Aaron Hagan [Thu, 21 Oct 2021 13:38:17 +0000 (09:38 -0400)]
Fix DS Layout in LoadStoreOpNone when using dynamic_rendering
It's in the wrong layout when being read from.
Components: Vulkan
VK-GL-CTS issue: 3236
Affected tests:
dEQP-VK.renderpass_with_dynamic_rendering.suballocation.load_store_op_none.*
Change-Id: I58ae176c6d179cfb262434b24937c5f278112ea6
Matthew Netsch [Fri, 22 Oct 2021 19:47:16 +0000 (15:47 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.3.0
Change-Id: I9f1e4eba89eb77636a4b58312efa86ada45b0945
Samuel Iglesias Gonsálvez [Fri, 1 Oct 2021 06:18:08 +0000 (08:18 +0200)]
Update drivers ids
From last version of Vulkan-Docs.
Components: Framework
VK-GL-CTS issue: 3173
Change-Id: I39aeb4f16d3bf19291a5a266595b215681ef2c83
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Matthew Netsch [Fri, 22 Oct 2021 18:23:39 +0000 (14:23 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7
Change-Id: I6f96b469518264e53fdd142f445187f8eb091c6d
Barry Ko [Wed, 27 Feb 2019 12:53:04 +0000 (20:53 +0800)]
Relax get_frame_timestamps
To prevent power side effect from legacy platform. We tend to relax the
deqp criteria from 3 vsyncs to 4 vsyncs. And verify timestamp 6 frames
ago.
Cherry pick from AOSP.
Original commit:
https://android.googlesource.com/platform/external/deqp/+/
db1f80ba923b9e89177562203c21192f8e7340c3
https://android-review.googlesource.com/c/platform/external/deqp/+/918914
Submitted on behalf of a third-party: barry.ko@mediatek.com
Component: EGL
Affected tests:
dEQP-EGL.functional.get_frame_timestamps.*
BUG:
123203074
Change-Id: I214b410d2ab8f0d6d4b3f9d6752c1c1d756cc28f
Chris Forbes [Wed, 13 Feb 2019 22:59:20 +0000 (14:59 -0800)]
Make renderdoc build-time integration optional
In some environments (Android Q CTS), renderdoc headers will not be
available.
Cherry pick from AOSP.
Original commit:
https://android.googlesource.com/platform/external/deqp/+/
7e79367bd8fa1430c0ea161a36f62d7b8edc6803
Components: Vulkan, Framework
Bug: b/
123431604
Change-Id: I7d4e36074e4f89a81f55dde83a217920653ad173
Matthew Netsch [Fri, 22 Oct 2021 17:07:14 +0000 (13:07 -0400)]
Merges vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6
Change-Id: I1cbb115f9efbb2df6de81b0b133723b15e367427
Matthew Netsch [Fri, 22 Oct 2021 17:01:32 +0000 (13:01 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5
Change-Id: I6873657c378b02b383da36905a31886806519459
Matthew Netsch [Fri, 22 Oct 2021 17:01:27 +0000 (13:01 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.3 into vk-gl-cts/vulkan-cts-1.2.4
Change-Id: I67159d811b780fc975c41a18043c177bceba6da7
Mika Väinölä [Wed, 13 Oct 2021 07:51:57 +0000 (10:51 +0300)]
Fix zx_handle_t typedef
Change from signed to unsigned 32-bit integer to match zircon/types.h.
Components: Framework, Vulkan
VK-GL-CTS issue: 3214
Change-Id: Ic887e90adc1583cfcd1518e21831d330e0f89b17
Jason Ekstrand [Mon, 21 Jun 2021 20:35:36 +0000 (15:35 -0500)]
Fix tessellation SPIR-V tests use of Restrict
Their usage of Restrict violates the SPIR-V spec which requires that it
only got on variables, not types or struct members.
Affected tests:
dEQP-VK.tessellation.*
Components: Vulkan
VK-GL-CTS issue: 2975
Change-Id: Ifbde1b493bf56d1ee6454a4ab54e98bcd05f839b
Mohd Faisal [Wed, 26 Aug 2020 10:51:25 +0000 (11:51 +0100)]
Allowing spirv1.4 shaders compilation
Components: Vulkan
VK-GL-CTS Issue: 2560
Affects: dEQP-VK.spirv_assembly.instruction.spirv1p4.*
Change-Id: Icbfe3ad623da98968eee401d08df5ed6b3b2db1e
(cherry picked from commit
c96d524d079fdada26b3548effd1539174415be6)
Samuel Iglesias Gonsálvez [Thu, 7 Oct 2021 09:29:38 +0000 (11:29 +0200)]
Fix VkClearRect sizes for vkCmdClearAttachment()
The test was doing out-of-bounds clears of attachments.
Components: Vulkan
VK-GL-CTS issue: 3174
Affected tests:
dEQP-VK.dynamic_rendering.*
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Change-Id: I6cbf82efdea824cc07aeaddf814368da5b8b5848
Piotr Byszewski [Fri, 1 Oct 2021 11:49:45 +0000 (13:49 +0200)]
SpirV 1.6: Test Nontemporal image operand
This change extends existing tests to cover cases where
new Nontempporal image operand is used with:
OpImageFetch, OpImageRead,
OpImageSampleExplicitLod,OpImageWrite,
OpImageSparseSampleExplicitLod, OpImageSparseFetch,
OpImageSparseRead and OpimageSparseGather.
VK-GL-CTS issue: 3127
Components: Vulkan
Affects:
dEQP-VK.spirv_assembly.instruction.compute.image_sampler.*
dEQP-VK.sparse_resources.shader_intrinsics.*
dEQP-VK.image.extend_operands_spirv1p4.*
dEQP-VK.image.nontemporal_operand.*
Change-Id: I1ba06b7e61e5d2512f9e61cfa14c96f45e1cedb0
Piers Daniell [Wed, 13 Oct 2021 03:44:04 +0000 (21:44 -0600)]
Partial revert of CL 8054: Fix invalid layout transitions in some dynamic rendering tests
This image barrier change was wrong. I thought this was a duplicate
barrier for the color attachment image, but it was actually
one for the depth/stencil attachment image.
Affects:
dEQP-VK.multiview.dynamic_rendering.depth.*
dEQP-VK.multiview.dynamic_rendering.stencil.*
Components: Vulkan
VK-GL-CTS issue: 3164
Change-Id: Id65e226eaec2c158289d684a0a997d8a929a5f45
Matthew Netsch [Fri, 15 Oct 2021 21:28:06 +0000 (17:28 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.3.0
Change-Id: Icb78e9ab972d7942276a0f12b6bcb55ceb2878c7
Samuel Iglesias Gonsálvez [Fri, 8 Oct 2021 06:38:23 +0000 (08:38 +0200)]
Add check for shaderImageGatherExtended
There were validation errors in the tests when running them
through devsim layer and vulkan 1.0 spec minimum profile:
"VUID-RuntimeSpirv-OpImage-06376(ERROR / SPEC): msgNum:
70978435 -
Validation Error: [ VUID-RuntimeSpirv-OpImage-06376 ] Object 0: handle
= 0x66d3840, type = VK_OBJECT_TYPE_DEVICE; | MessageID = 0x43b0b83 |
vkCreateShaderModule(): Shader uses OpImageGather with offset (-8)
less than VkPhysicalDeviceLimits::minTexelGatherOffset (0). The Vulkan
spec states: If an OpImage*Gather operation has an image operand of
Offset, ConstOffset, or ConstOffsets the offset value must be greater
than or equal to minTexelGatherOffset"
Components: Vulkan
VK-GL-CTS issue: 3186
Affects:
dEQP-VK.glsl.texture_gather.offset.*
Change-Id: I1c58d751a86dbe7a1a285a32daae19955b5b4582
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Mohd Faisal [Wed, 13 Oct 2021 14:09:01 +0000 (15:09 +0100)]
Fix missing self-subpass dependency
Self-subpass dependency required for read-after-write operation, in test
cases where the attachment is written using load-op-load at the start
of subpass and then the values are read inside the fragment shader.
Affects:
dEQP-
VK.renderpass.suballocation.subpass_dependencies.separate_channels.*
dEQP-
VK.renderpass2.suballocation.subpass_dependencies.separate_channels.*
Components: Vulkan
VK-GL-CTS issue: 3216
Change-Id: I6c72b81dccfb7526ec4607b255ff4872b13c4d22
Samuel Iglesias Gonsálvez [Wed, 13 Oct 2021 08:59:26 +0000 (10:59 +0200)]
Fix workgroup_memory_explicit_layout tests SPIR-V validation errors
Components: Vulkan
VK-GL-CTS issue: 3195
Affects:
dEQP-VK.compute.workgroup_memory_explicit_layout.*
Change-Id: I2ac798965d9a3518bc45cfe52e43f34facf4ccf4
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Erik Faye-Lund [Wed, 11 Aug 2021 14:18:37 +0000 (16:18 +0200)]
Verify line-stipple both with and without carry-over
The Vulkan spec specifies that carrying over the fractional parts of a
line is optional. Let's test both with and without, so we accept either.
Components: Vulkan, Framework
Public Issue: 280
Affected Tests: dEQP-VK.rasterization.primitives.static_stipple.*
dEQP-VK.rasterization.primitives_multisample_4_bit.*
Change-Id: Iaae3397ebfa1db7fbccefd479bd98c1ad8471ff8
Matthew Netsch [Fri, 15 Oct 2021 20:24:58 +0000 (16:24 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7
Change-Id: I02d8c267d85372b375fcdcb1df69f2c6dc97796c
Joe M. Kniss [Fri, 16 Nov 2018 16:59:30 +0000 (08:59 -0800)]
Add memoryBarrier before buffer map verify
OpenGL ES 3.2 Specification Section 7.11.2:
Shader Memory Access Synchronization ...
"Explicit syncronization is required to ensure the effects of buffer and
texture data stores performed by shaders will be visible to subsequent
operations using the same objects..."
Many of the ES3.1 compute tests lack a memory barrier before mapping and
verifiying buffer contents. This can lead to flaky test failures when
the buffer has not been completely written to by the dispatchCompute
command. Specifically, chromeos and it's android arc++ has seen
consistent, yet flaky failures in es31fSynchronizationTests.inter_invocation.
It is unclear why this test should fail when similar tests, which also
lack explicit synchronization, do not.
Cherry pick from AOSP.
Original commit:
https://android.googlesource.com/platform/external/deqp/+/
30c710319864b9e952daaaaf098df20845c2f70d
https://android-review.googlesource.com/c/platform/external/deqp/+/854112
TEST=chromeos caroline-arcnext (skylake)
Affects: dEQP-GLES31.functional.synchronization.*
Components: AOSP
Google Bug: b/
85161472
Change-Id: I5093e7e206996e7c13c8c3615ee01ebbadc42f58
Signed-off-by: Joe M. Kniss <djmk@google.com>
Paul Thomson [Thu, 7 Oct 2021 17:37:55 +0000 (18:37 +0100)]
Update Android test lists and Android files
Update Android test lists and other files to match AOSP
android12-tests-dev. Some tests are excluded from Android test lists and
some are moved between different Android test lists, but the tests
themselves do not change.
Components: AOSP
Change-Id: I28fdab5eb0b3fa75604f525a34fe843091836e75
Paul Thomson [Thu, 7 Oct 2021 17:37:55 +0000 (18:37 +0100)]
Update Android test lists and Android files
Update Android test lists and other files to match AOSP
android12-tests-dev. Some tests are excluded from Android test lists and
some are moved between different Android test lists, but the tests
themselves do not change.
Components: AOSP
Change-Id: I28fdab5eb0b3fa75604f525a34fe843091836e75
James Darpinian [Tue, 18 Dec 2018 00:46:59 +0000 (16:46 -0800)]
EXT_multisampled_render_to_texture test
Detects a bug where ReadPixels fails on multisampled textures.
See http://crbug.com/890002
Cherry pick from AOSP.
Original commit:
https://android.googlesource.com/platform/external/deqp/+/
accf1be52e32153c757cd84864607ee9c47dfe9a
https://android-review.googlesource.com/c/platform/external/deqp/+/796566
Submitted on behalf of a third-party: jdarpinian@chromium.org
Components: OpenGL
New test:
dEQP-GLES2.functional.multisampled_render_to_texture.readpixels
Bug: chromium:890002
Change-Id: I6e5e5d1fa3aa15daee69271339962a0b3ffc1927
sriharsha [Wed, 29 May 2019 05:02:22 +0000 (10:32 +0530)]
Increase tolearance for dithering tests
For small dimension render targets, gradient tests
require more tolerance, as variance in adjacent pixels
will be more compared to bigger render targets.
Cherry pick from AOSP.
Original commit:
https://android.googlesource.com/platform/external/deqp/+/
8dc015fa12e41a380b81c1b8fcc63a04bf34311a
https://android-review.googlesource.com/c/platform/external/deqp/+/971506
Submitted on behalf of a third-party: quic_spvasu@quicinc.com
Components: OpenGL
Affected tests:
dEQP-GLES3.functional.dither.*
Bug:
132146940
Change-Id: Ie48e119d164acf5763aa7956c3ff91da20ff9f38
Pierre Couillaud [Wed, 31 Oct 2018 11:57:40 +0000 (04:57 -0700)]
Add watchdog touches to descriptor_pool test
On some low-end devices, this test would hit the dEQP watchdog. Add some
watchdog touches to avoid this.
Cherry pick.
Original commit:
123de98482b779ca242f9498701508b12e9a5f01
https://gerrit.khronos.org/c/vk-gl-cts/+/3197
Affected tests:
dEQP-VK.api.descriptor_pool.*
Google Bug: b/
116847904
Component: Vulkan
Signed-off-by: Pierre Couillaud <pierre@broadcom.com>
VK-GL-CTS Issue: 1464
Change-Id: I18181aa274ab1dc191b5f0acbb58b41465b5cdad
Ricardo Garcia [Wed, 13 Oct 2021 09:47:43 +0000 (11:47 +0200)]
Check fragment shader stores in frag shader side effect tests
Fragment shader side effect tests were using storage buffer store
operations in fragment shaders without checking support for the
fragmentStoresAndAtomics feature.
In addition, the tests were verifying the implementation supported the
color attachment and transfer features for the VK_FORMAT_R8G8B8A8_UNORM
format, but they are mandatory by the spec. The checks can be safely
removed.
Affected tests:
dEQP-VK.rasterization.frag_side_effects.*
Components: Vulkan
VK-GL-CTS issue: 3208
Change-Id: I5d28b18dd6aedbbfb1ea3ce4b40b3ba82737c27b
Ricardo Garcia [Wed, 13 Oct 2021 08:23:04 +0000 (10:23 +0200)]
Use the correct view type in no_position tests
Multilayer no_position tests were creating image views with multiple
layers and the VK_IMAGE_VIEW_TYPE_2D type. When the number of layers is
greater than one, the type should be VK_IMAGE_VIEW_TYPE_2D_ARRAY.
Affected tests:
dEQP-VK.pipeline.no_position.*
Components: Vulkan
VK-GL-CTS issue: 3200
Change-Id: I9d24062c4f1ac80f12ac5dd1fe0e454ffade2f32
Ricardo Garcia [Wed, 13 Oct 2021 08:57:20 +0000 (10:57 +0200)]
Check fragment shader stores in robustness tests
Some robustness tests were using fragment shader stores without checking
for the fragmentStoresAndAtomics feature.
Affected tests:
dEQP-VK.robustness.*.frag
Components: Vulkan
VK-GL-CTS issue: 3209
Change-Id: I8032bd19f7de4f2d87d642542c2b7aaec20965b8
Ricardo Garcia [Wed, 13 Oct 2021 08:54:44 +0000 (10:54 +0200)]
Check cube array support in robustness tests
Some robustness tests were using VK_IMAGE_VIEW_TYPE_CUBE_ARRAY image
views without checking support for the imageCubeArray feature.
Affected tests:
dEQP-VK.robustness.*cube_array*
Components: Vulkan
VK-GL-CTS issue: 3210
Change-Id: Iafd363a6423251413a1093e15ae94aa50850af32
Matthew Netsch [Fri, 15 Oct 2021 17:57:59 +0000 (13:57 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6
Change-Id: I64019cfcae5bcedae55f3e343bc94f51b62aa877
Matthew Netsch [Fri, 15 Oct 2021 17:57:51 +0000 (13:57 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5
Change-Id: I0abe7100c16a273d62463b916957255150d344d0
Matthew Netsch [Fri, 15 Oct 2021 17:57:39 +0000 (13:57 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.3 into vk-gl-cts/vulkan-cts-1.2.4
Change-Id: Ib23f14c29d67945839ab3763968963d7dcc91464
Samuel Iglesias Gonsálvez [Fri, 8 Oct 2021 13:26:21 +0000 (15:26 +0200)]
Fix validation errors with unnormalized coordinates samplers
Components: Vulkan
VK-GL-CTS issue: 3190
Affects:
dEQP-VK.pipeline.sampler.*unnormalized*
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Change-Id: I1dde344fe39289b945a434eab7276dc58deab881
Graeme Leese [Thu, 7 Oct 2021 17:15:51 +0000 (18:15 +0100)]
Fix descriptor indexing test dependency
The UBO and SSBO non-uniform indexing tests use runtimeDescriptorArray,
so this needs to be included in the checkSupport condition.
Components: Vulkan
Affects: dEQP-VK.ubo.*
dEQP-VK.ssbo.*
Change-Id: I54edb2d06135a8d1d549a42ac97014dabc80c9dd
Samuel Iglesias Gonsálvez [Wed, 2 Jun 2021 14:41:30 +0000 (16:41 +0200)]
Check resource limits properly
The tests were not checking the resource limits per stage when
creating the shaders, which could create potential issues on
some drivers. Specially the random generated tests.
Affected tests:
dEQP-VK.ssbo.layout.*
Components: Vulkan
VK-GL-CTS issue: 2953
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Change-Id: I7509dce38b849a0dd26f9e47d7f74e637930d897
Matthew Netsch [Fri, 15 Oct 2021 16:47:17 +0000 (12:47 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3
Change-Id: Id485bc8ae929179e7786039966a233dd6fa57cdc
Vihanakangas [Mon, 4 Oct 2021 10:52:27 +0000 (13:52 +0300)]
Move anisotropic filtering support check to a function
Support check for anisotropic filtering was done inside of iterate.
This change moves that check inside of a function for consistency
and performance.
Affects:
dEQP-VK.texture.filtering_anisotropy.*
Components: Vulkan
VK-GL-CTS issue: 3068
Change-Id: I3a1c0a1acff96b6f6b291bb53d3d9dc0bd122c17
(cherry picked from commit
522f4ba64beee8828030320591d20faa027a8e0a)
Piotr Byszewski [Mon, 4 Oct 2021 10:13:36 +0000 (12:13 +0200)]
Fix extended_dynamic_state2 tests
Before this change alias assignment was only done for
promoted extensions. With this change search for aliases
is done for all extensions listed in extensions_data.txt.
This mainly fixes issue with tests for partially promoted
VK_EXT_extended_dynamic_state2. Three entrypoints that
were not promoted to vk1.3 must still use EXT when
vk1.2 implementation is tested.
This change also removes one duplicate from extensions_data.txt.
Components: Vulkan, Framework
VK-GL-CTS Issue: 3159
Affects: dEQP-VK.pipeline.extended_dynamic_state.*
Change-Id: I552764b3559a6f78d84a2f764f0cc5250c52f0ed
James Fitzpatrick [Tue, 5 Oct 2021 09:12:46 +0000 (10:12 +0100)]
Fix incorrect shader output format
For test variants that write to a stencil buffer, the final attachment
will be R8_UINT. For these tests ensure the fragment shader is
outputing to this attachment as a uint.
Affects: dEQP-VK.synchronization2.none_stage.*
Components: Vulkan
VK-GL-CTS issue: 3177
Change-Id: Ia7e185c8c8486c8b0e76e11038e0a5108770f21f
James Fitzpatrick [Wed, 6 Oct 2021 13:22:03 +0000 (14:22 +0100)]
Increase tolerance for matrix precision tests
On hardware with a RTZ rounding mode the tolerance required in these
tests was too low and valid implementations of the functions would
fail the tests.
Increase the tolerance allowed to allow RTZ hardware to pass these
tests.
Affects:
dEQP-VK.glsl.builtin.precision*.determinant.*
dEQP-VK.glsl.builtin.precision*.inverse.*
Components: Vulkan
VK-GL-CTS issue: 3182
Change-Id: I83845aec8faa01aec0820852219b6b8c4326da4e
Matthew Netsch [Fri, 8 Oct 2021 18:42:39 +0000 (14:42 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.3.0
Change-Id: Icd84bfefac281e224692dd83c2ac6dff87f9f43c
Matthew Netsch [Fri, 8 Oct 2021 18:02:50 +0000 (14:02 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7
Change-Id: I2eaed7035330ecb962466f205b3d3dd80a311d4b
Lars-Ivar Hesselberg Simonsen [Tue, 28 Sep 2021 08:20:00 +0000 (10:20 +0200)]
Fix missing barrier in FSR pixel consistency tests
The PixelConsistency tests of Fragment Shading Rate did not have a
pipeline barrier between the calls to vkCmdClearColorImage for the
render targets and the subsequent renderpass. This caused a race between
the two.
This commit adds the missing pipeline barriers.
In addition, it converts the preImageBarrier from two calls to
vkCmdPipelineBarrier with a single barrier to a single call with two
barriers.
Components: Vulkan
Affects: dEQP-VK.fragment_shading_rate.pixel_consistency.*
VK-GL-CTS Issue: 3158
Change-Id: I14b5de4fb02593b0a384b0f36892543ee2c57222
Ricardo Garcia [Mon, 27 Sep 2021 09:05:12 +0000 (11:05 +0200)]
Do not enable anisotropic filter in exact sampling tests
The anisotropic filter may introduce some extra imprecision in some
implementations. This change also makes these tests have the same
behavior in all implementations.
Affected tests:
dEQP-VK.pipeline.sampler.exact_sampling.*
Components: Vulkan
VK-GL-CTS issue: 3152
Change-Id: I0b4e1ad2c46210568176c5bd5504678edb3dbc73
Paul Thomson [Mon, 20 Sep 2021 15:53:45 +0000 (16:53 +0100)]
Reduce resolution of some graphicsfuzz tests
These tests have a lot of shader computation, which may exceed a 30
second watchdog time limit. Reducing the resolution should allow them to
complete quickly on low-end devices.
Affects:
dEQP-VK.graphicsfuzz.nested-for-loops-switch-fallthrough
dEQP-VK.graphicsfuzz.spv-access-chains
dEQP-VK.graphicsfuzz.spv-copy-object
dEQP-VK.graphicsfuzz.spv-dead-break-and-unroll
dEQP-VK.graphicsfuzz.spv-stable-maze-flatten-copy-composite
Components: Vulkan
Change-Id: I4ca78ae1beafb3d892d437568f10cc816ae79df2
Matthew Netsch [Fri, 8 Oct 2021 15:42:39 +0000 (11:42 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6
Change-Id: I1d4bf5c3a9cbd2f7d8e5bdce8645065661be1e9c
Marcin Kańtoch [Tue, 31 Aug 2021 11:08:45 +0000 (13:08 +0200)]
Fix incorrect computation of pointer size
CTS always fills the array of pointers with a stride of 64-bits.
So, for 32-bits applications, each two 32-bits has one valid pointer
to instance of bottom-level acceleration structure.
To correct this the pointer size should be calculated as
sizeof(VkDeviceOrHostAddressConstKHR::hostAddress) for host built
top-level acceleration structure and
sizeof(VkDeviceOrHostAddressConstKHR::deviceAddress) for device built
top-level acceleration structure.
VK-GL-CTS issue: 3084
Components: Vulkan
Affects:
dEQP-VK.ray*instances_aop*
Change-Id: I8c412a5a5d2ce3f1f687795768d6d83ad4913892
Matthew Netsch [Thu, 7 Oct 2021 20:56:05 +0000 (16:56 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5
Change-Id: I911747cbf1726e161c457fa6ae666148ff0b36c7
Matthew Netsch [Thu, 7 Oct 2021 20:55:56 +0000 (16:55 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.3 into vk-gl-cts/vulkan-cts-1.2.4
Change-Id: I6994a2ac890d752906a8f6464a9cb7819e22a55f
Samuel Iglesias Gonsálvez [Tue, 28 Sep 2021 09:05:30 +0000 (11:05 +0200)]
Fix crashes on depth stencil resolve tests with separate layouts
Mesa drivers were crashing in depth stencil resolve tests when
used with separate layouts.
The issue was that if layout only specifies the layout of the
depth aspect of the attachment, the layout of the stencil aspect
is specified by the stencilLayout member of a
VkAttachmentReferenceStencilLayout structure included in the
pNext chain. We were not setting it.
Affected tests:
dEQP-VK.renderpass2.depth_stencil_resolve.*separate_layouts*
Components: Vulkan
VK-GL-CTS issue: 3157
Change-Id: I72f8fdba88717ab0442501911fb26a7469da65d8
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Samuel Iglesias Gonsálvez [Thu, 15 Jul 2021 09:15:32 +0000 (11:15 +0200)]
dEQP-VK.subgroups.*ray_tracing* missing extension check
Components: Vulkan
VK-GL-CTS issue: 3003
Affected:
dEQP-VK.subgroups.*ray_tracing*
Change-Id: I25719d00c6d0180c03726d73497727754f9539d2
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
(cherry picked from commit
a63763167972c24e5a2167cb78dc009eb161a6d4)
Piotr Byszewski [Thu, 30 Sep 2021 13:12:20 +0000 (15:12 +0200)]
Update dynamic rendering tests to separate formats
Depth and stencil formats were split in VK_KHR_dynamic_rendering.
VK-GL-CTS issue: 3161
Components: Vulkan
Affects:
dEQP-VK.draw_with_dynamic_rendering.*
dEQP-VK.multiview.dynamic_rendering.*
dEQP-VK.fragment_shading_rate_with_dynamic_rendering.*
dEQP-VK.renderpass_with_dynamic_rendering.*
dEQP-VK.dynamic_rendering.*
Change-Id: Id28ccb278ad6cc52b733d31076343874762d8c39
Slawomir Cygan [Tue, 28 Sep 2021 15:37:46 +0000 (17:37 +0200)]
Fix stencil data unpacking in dEQP-VK.fragment_operations test
The test copies only one aspect (depth or stencil, not both) of the image
to the buffer, so host unpacking code should not use combinded DS format
- instead S8_UIN format should be used in stencil tests.
The problem occurs in stencil tests, on implementations not supporing
S8_UINT format, where D+S format is used.
Components: Vulkan
VK-GL-CTS Issue: 3160
Affects: dEQP-VK.fragment_operations.*stencil
Change-Id: I27b69c02ebf08fa19f512cdd786ebe596d91cfeb
Piers Daniell [Wed, 29 Sep 2021 14:04:22 +0000 (08:04 -0600)]
Fix invalid layout transitions in some dynamic rendering tests
Some of the dynamic rendering tests do an image layout transition
from uninitialized when the image is already in use. This can lead
to corruption of the image content.
This CL removes an unnecessary barrier from the
dEQP-VK.multiview.dynamic_rendering.readback* tests and fixes
a barrier with the dEQP-VK.multiview.dynamic_rendering.depth.*
and dEQP-VK.multiview.dynamic_rendering.stencil.* tests.
Affects:
dEQP-VK.multiview.dynamic_rendering.depth.*
dEQP-VK.multiview.dynamic_rendering.stencil.*
dEQP-VK.multiview.dynamic_rendering.readback*
Components: Vulkan
VK-GL-CTS issue: 3164
Change-Id: Ib0847839528909337001d32bbde59c7e979ec9e9
Slawomir Cygan [Wed, 29 Sep 2021 16:26:49 +0000 (18:26 +0200)]
Do not chain dynamic render pass inheritance info in dEQP-VK.renderpass
This fixes the tests that do not use, enable or require dynamic rendering
to not chain the VkCommandBufferInheritanceRenderingInfoKHR structure,
which is only for inheritinh the dynamic renderpasses.
Affects: dEQP-VK.renderpass*
VK-GL-CTS Issue: 3165
Components: Vulkan
Change-Id: Idb8fed2bcd6a6bf664cf5516a664b8beaa94484a
Aaron Hagan [Mon, 27 Sep 2021 19:22:47 +0000 (15:22 -0400)]
Wrong format in load_store_op_none test.
One of the load_store_op_none tests is currently using the incorrect
VK_FORMAT_R8G8B8A8_UNORM format for the color attachments.
It should be using VK_FORMAT_R8G8B8A8_UINT instead.
VK-GL-CTS issue: 3154
Components: Vulkan
Affected tests:
dEQP-VK.renderpass_with_dynamic_rendering.suballocation.load_store_op_none.color_load_op_none_store_op_none_resolve
Change-Id: I7dd8b4da05fef58d91a8913ac49912ba683f033d
Matthew Netsch [Fri, 1 Oct 2021 22:05:39 +0000 (18:05 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.3.0
Change-Id: Icd53be0bdc32472d88bd464dffb0ee3a8ab3677e
Erik Faye-Lund [Wed, 7 Jul 2021 16:39:28 +0000 (18:39 +0200)]
Fix rectangular line tests
When we're testing against rectangular lines, we need to generate the
rectangular line geometry as well, otherwise we'll test against
parallelogram lines instead.
Components: Vulkan, Framework
Public issue: 274
Affected Tests: dEQP-VK.rasterization.primitives.*.rectangular_line_*
Change-Id: I514646e94d2faf1e996950fd6b3a738797f84ff0
Matthew Netsch [Fri, 1 Oct 2021 21:06:12 +0000 (17:06 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7
Change-Id: I015e15ff85941f0a8756bd4b13e93fcd9161d62e
Samuel Iglesias Gonsálvez [Fri, 24 Sep 2021 10:55:04 +0000 (12:55 +0200)]
Fix dEQP-VK.glsl.opaque_type_indexing.ssbo*_compute validation error
Tests were not checking maxPerStageDescriptorStorageBuffers limits
in the compute shaders, which use more (6) than the minimum limit
set by the spec for that (4).
Components: Vulkan
VK-GL-CTS issue: 3137
Affected tests:
dEQP-VK.glsl.opaque_type_indexing.ssbo*_compute
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Change-Id: I016fb5024d637e36869ae7e3f78467c91d0e8817
Juha Heiskanen [Fri, 24 Sep 2021 10:12:28 +0000 (13:12 +0300)]
Fix dEQP-VK.image.extended_usage_bit.texture_* validation errors
Shader requires fragmentStoresAndAtomics but there wasn't any support
check for this physical device feature.
Components: Vulkan
VK-GL-CTS Issue: 3144
Affects:
dEQP-VK.image.extended_usage_bit.texture_read.*
dEQP-VK.image.extended_usage_bit.texture_write.*
Change-Id: I87f69256501519fe524159f02eb14ee6ac0e0092
Ricardo Garcia [Tue, 21 Sep 2021 10:27:52 +0000 (12:27 +0200)]
Check depth bias clamp feature when needed
The inverted depth range tests were using the depth bias clamp feature
with a nonzero depth bias clamp value without checking if the feature
was available in the device.
Affected tests:
dEQP-VK.draw.inverted_depth_ranges.*
Components: Vulkan
VK-GL-CTS issue: 3128
Change-Id: I471235a0da38eb4c244e87e0d5851b8cb810f990
David Neto [Tue, 21 Sep 2021 21:25:22 +0000 (17:25 -0400)]
Add barrier between the atomic-compare-exchanges
Each of the second atomic-compare-exchange instructions relies on fact
that all the invocations have executed the first atomic-compare-exchange
instruction. So insert a workgroup barrier between the two.
Affects:
dEQP-VK.spirv_assembly.instruction.spirv1p4.entrypoint.comp_workgroup_entry_point
Components: Vulkan
VK-GL-CTS issue: 3120
Change-Id: I95f897ceebcea8fdd5d6bda885a701be20d07ab1
Matthew Netsch [Fri, 1 Oct 2021 19:33:00 +0000 (15:33 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6
Change-Id: I16fde981fb8a0d98d4ed0522cb69a429a68545d0
Matthew Netsch [Fri, 1 Oct 2021 19:32:56 +0000 (15:32 -0400)]
Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5
Change-Id: I133af15a49b220d94c26f9db70e7830fbe167b72
Alexander Galazin [Fri, 18 Jun 2021 08:17:32 +0000 (10:17 +0200)]
Allow the use of the standard types
Removes the script that enforces the use of
the de-wrappers instead of the standard types.
VK-GL-CTS issue: 2963
Components: Framework
Change-Id: I5e2ff38a2b2cbc63a3c756a1e0260833ecd9381e
(cherry picked from commit
ce1ae17c4785a8a57214a16fac46e28832956fa9)
Samuel Iglesias Gonsálvez [Tue, 28 Sep 2021 12:32:25 +0000 (14:32 +0200)]
Fix dEQP-VK.image.mutable.*srgb* validation errors
The tests were not checking that the image format was
supported before creating the image.
Components: Vulkan
VK-GL-CTS issue: 3143
Affected tests:
dEQP-VK.image.mutable.*
Change-Id: Ia42ccc815456ac09c4ffd44c63a0c65db8a9f1ec
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Piotr Byszewski [Tue, 21 Sep 2021 16:22:35 +0000 (18:22 +0200)]
Fix Vulkan 1.3 required features
vulkanMemoryModelAvailabilityVisibilityChains feature is now
optional in Vulkan 1.3
VK-GL-CTS issue: 3116
Components: Vulkan
Affects:
dEQP-VK.api.info.vulkan1p3.feature_extensions_consistency
Change-Id: If9d6e32f96c7053bdc0908ba361c1603498ae0df
Graeme Leese [Wed, 22 Sep 2021 13:16:25 +0000 (14:16 +0100)]
Allow 1.3.0.0 as a valid conformance version
Components: Vulkan
VK-GL-CTS issue: 3146
Affects: dEQP-VK.api.driver_properties.conformance_version
Change-Id: Ic77a90e47f9d649ebb5b3a3344876bcbc54835b9