Fix dEQP-VK.image.sample_texture.* tests write out-of-bounds bug

Compute shader writing to the view is now made by using block size
instead of pixel count.

Components: Vulkan

VK-GL-CTS Issue: 3595

Affects: dEQP-VK.image.sample_texture.*

Change-Id: I50879fd7bb48f623e565c8972e7a517fb26f0ed0

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: Id90879713fa5388ab2a4ebdd6811b454b0675763

Fix extension promotions

VK_EXT_shader_image_atomic_int64 was promoted instead of
VK_KHR_shader_atomic_int64.
VK_KHR_external_semaphore lost its promotion.
VK_KHR_vulkan_memory_model was promoted to the wrong core version.

VK-GL-CTS issue: 3631

Affects:
dEQP-VK.*

Change-Id: I78712093945796adf4076e75f6f3e5c07a4ec832

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: Ie2a23e7c3811d6cebc0b7ace2351d5a4779bd7d0

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: I6ab43b2abea09721ad28bdba33565c7c0f0be5fd

Fix warnings with clang and GCC

Affected tests:
No test result should be affected by these changes.

Components: OpenGL, Vulkan
VK-GL-CTS issue: 3602

Change-Id: If2eb0736d296add732aaff53c80a705cc306d7a3

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: I2c051a776b06d23239eb3ce501a999088cd0c1de

Rework extensions_data and mandatory_features

Data from mandatory_features.txt was split to multiple json files
one per extension/vulkan_version. Those files now also include data
that was previously stored in extensions_data.txt .

VK-GL-CTS issue: 3197

Components: Framework, Vulkan

Affects:
dEQP-VK.*

Change-Id: I9756237829366ff4140dcf59dc7548a8f4328cb7

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: I69f0b9b087df9c33b26ddd014fc1a291fc50dbdd

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: I4a1b44642747f72803ff97dafd55f4ba3fabad9e

Use associated physical device from custom instance

Some of the API feature info tests created a custom instance,
but still used the default physical device. This is fixed by
choosing a physical device using the custom instance.

Comparison of VkPhysicalDeviceMemoryProperties was also
changed to only check the active memoryTypes and memoryHeaps
array elements instead of doing one large deMemCmp.

VK-GL-CTS Issue: 3590

Affects:

dEQP-VK.api.info.*

Components: Vulkan
Change-Id: I2378bcd1e97b7f85e533d34c5b5f6ac3e4fa9b93

Wait for deferred operations before obtaining RT pipeline handle

The ray tracing utils, when creating ray tracing pipelines and using
deferred operations, were using the VkPipeline pointer and handler
before the deferred operation had completed, which is against the spec.

Affected tests:
dEQP-VK.ray_tracing_pipeline.pipeline_library.configurations.multithreaded_compilation_dho.*
dEQP-VK.ray_tracing_pipeline.large_shader_set.cpu_ht*

Components: Vulkan
VK-GL-CTS issue: 3588

Change-Id: Ied8f443c01e99878180130963b8988852171b8a0

Fix subgroup size control failure logging

One of the failure messages was missing the message terminator, which
prevented it from ever being logged.

Affects:
dEQP-VK.subgroups.size_control*

Components: Vulkan
VK-GL-CTS issue: 3596

Change-Id: If649a11e0f6cf0f5e69e438d46b385b26e57dfe4

Parse test case result version when converting to XML

When parsing QPA files, extract the test case result version so it can
be later used to print the proper version when converting to XML files.

Components: Framework
VK-GL-CTS issue: 3573
VK-GL-CTS public issue: 287

Change-Id: I34d005621685b0ef31a7092ec7c40e6626bf20e3

Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5

Change-Id: I25f8f2f56ab07a9c3f04e8cfa54f2efc11bfd80a

Add queue ownership release in modifier tests

Queue ownership transfers require a release operation and an acquire
operation. Before this patch, this test only performed the acquire
operation.

Components: Vulkan

VK-GL-CTS issue: 3575

Affects: dEQP-VK.drm_format_modifiers.export_import.*
Change-Id: I2998d7928d1042d6a159e1173257ee551c1120a5

Fix incorrect layout transition in modifier tests

The queue family ownership acquire operation specifies a layout
transition from the pre-initialized layout. The image isn't in the
pre-initialized layout however, it's in the transfer dst optimal layout.
Align with the Vulkan spec, which says:

When performing a layout transition on an image subresource, the
old layout value must either equal the current layout of the
image subresource (at the time the transition executes), or else
be VK_IMAGE_LAYOUT_UNDEFINED (implying that the contents of the
image subresource need not be preserved).

Components: Vulkan

VK-GL-CTS issue: 3575

Affects: dEQP-VK.drm_format_modifiers.export_import.*
Change-Id: I76dbb54c699ed7d34e964de17f71ef01bc9e98a5

Fix lossy layout transitions in modifier tests

The test performs several layout transitions that use
VK_IMAGE_LAYOUT_UNDEFINED for the initial layout. The spec says that
these transitions cause images to become undefined. The test expects
defined data from these images however. Replace these initial layouts
with layouts that don't cause data loss.

Components: Vulkan

VK-GL-CTS issue: 3575

Affects: dEQP-VK.drm_format_modifiers.export_import.*
Change-Id: I5e57dd197d461fba800d50239803d525c5334ac4

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: I4713730ea2497ef56f0c939a5e73a6bdc0cdcb6e

Fix memory leaks when creating test hierarchy

The lack of a virtual destructor in one of the classes used as part of
some test parameters made createDescriptorValveMutableTests() leak
memory whenever it was called, in the moment the test hierarchy is being
destroyed. The leak will happen not only when running mutable descriptor
test, but when the binding_model test group needs to be generated for
any reason, including running tests in adjacent groups.

Hence, the change has no specific list of affected tests. At the same
time, no test result should be affected by the change.

Components: Vulkan
VK-GL-CTS issue: 3583

Change-Id: I7e847b5927359cef27f9992cd53991dd1a7cb781

Use proper image tiling in synchronization test support checks

Some test variants use LINEAR instead of OPTIMAL tiling and the proper
tiling mode features need to be checked.

Affected tests:
dEQP-VK.synchronization.*

Components: Vulkan
VK-GL-CTS issue: 3572

Change-Id: I16042ace7fb3a2f93b1503e539a5bbe59e03ae14

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: I5e73a149353915a09697633abab737ee8a38fdca

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: I4624140325b398fc901142f54425375b5ffa3d66

Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5

Change-Id: Ibcfcfa0cf5ad247b4c80017ad298d15a5139868f

modifiers: Use dedicated memory consistently

When a dedicated allocation device memory object
is exported, it must be imported as a dedicated
allocation created with a compatible resource as
well.  This change makes the DRM format modifier
tests self-consistent in that regard.

Components: Vulkan

Affected tests: dEQP-VK.drm_format_modifiers.export_import.*

VK-GL-CTS issue: 3544

Change-Id: I93a35180b9a5a1b98e0d4fa1f2934258134cef1b

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: I61b2991c59749bd8fe611add1809f596035a1027

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: Ie50f49d88d9a088bd86f6a659349557d5ad50cea

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: Iefb6b35452197d43db9dfc52b689476d2a81b8dd

Fix Zlib depedency

Use newer version of Zlib as older is no longer available to download from the zlib.net page

Components: Framework

Affects:

None

VK-GL-CTS Issue: 3587

Change-Id: Ie166dc897b0d1318c1ae73434bfa19687b579691

Increase tolerance in mat3 determinant compute test

The determinant operation does not have strict precision requirements
and a recent change in Mesa broke this test for RADV. A slight increase
in tolerance fixes it.

Affected tests:
dEQP-VK.glsl.builtin.precision_double.determinant.compute.mat3

VK-GL-CTS issue: 3571
Components: Vulkan

Change-Id: I5185bc4d45d848cbad11798f743ae65fbb798657

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: Ie55187998eab0ae965e731fc98511b86c842704e

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: I81f233780e77ccfb79e9b1b9d982c9406a9ca268

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: I8a313ea2aebdd5163a689b2cc31161e5963172a2

Check for visibility in external_memory_host test

The external_memory_host synchronization test uses mapMemory when
validating the contents of the external memory, but does not ensure that
the memory is host-visible. When choosing a memory type for the
host-visible memory, additionally check the HOST_VISIBLE flag.

Affects:

dEQP-VK.memory.external_memory_host.synchronization.synchronization

Components: Vulkan

VK-GL-CTS issue: 3547

Change-Id: I0cff8f42debedd01db1760881a6f5f9481ea39bf

Fix dEQP-VK.subgroups.size_control.compute.required_subgroup_size_*_require_full_subgroups

The tests were testing full subgroups with four workgroup sizes:

 - Nx1x1
 - 1xNx1
 - 1x1xN
 - getLocalSizes() which returns a maximally large workgroup size

However, the Vulkan spec says:

    "If a VkPipelineShaderStageRequiredSubgroupSizeCreateInfo structure
    is included in the pNext chain, and flags has the
    VK_PIPELINE_SHADER_STAGE_CREATE_REQUIRE_FULL_SUBGROUPS_BIT flag set,
    the local workgroup size in the X dimension of the pipeline must be
    a multiple of
    VkPipelineShaderStageRequiredSubgroupSizeCreateInfo::requiredSubgroupSize

This means that the middle two subgroup sizes are invalid for use with
VK_PIPELINE_SHADER_STAGE_CREATE_REQUIRE_FULL_SUBGROUPS_BIT.

Components: Vulkan

Affects:
 - dEQP-VK.subgroups.size_control.compute.required_subgroup_size_max_require_full_subgroups
 - dEQP-VK.subgroups.size_control.compute.required_subgroup_size_min_require_full_subgroups

Change-Id: I46f2547e1a50f78cbed6695f96448f463ca57230

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: I2280d6ce320738b91994d4603600c38d5d1ba4ee

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: If390df3efc02bff47f240c62e35d23e78c537b4c

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: I31b3e7ec830aa10aa3c06ef61972a4e9c48dffc2

Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5

Change-Id: I010f996f4b726f1f79e9bbe9b9f3e96131021f00

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

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

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: Ib6194cc5034a2dada976ef494cf4c23fe997555f

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

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

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: I589df8409840b42f6c683dda57616b3ee8359128

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: I3cf7433d510e0966f4f4d96971519aa849cecda0

Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5

Change-Id: I8e72cdeaa8eed150a5548dd30871eccb09f31d5c

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

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: I37050cfb18cd34bd869b26531623f3b881973a3b

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

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: I3e993e882a340347eea3582c1ba4feb3393b58fc

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: I3ae2bc92916e5c4e2b525a085ba763d4c4b124d1

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

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

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

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

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

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

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: If3af09eda7b56dbfa484de49f06401ff53fda906

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

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: Iba2783e1e1e2bcb839c075dcf4d2b1dbcfbdf43f

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

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: I4adbb2e30086c26c8ad3312a6086e267c2840d42

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

Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5

Change-Id: Ib8646342ae0c6c9df3c7f2fa78c24643c10d5e96

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

Delay initializing fragment_shading_rate tests

When the shading rate tests were renamed, they turned a delayed call
to createChildren into an immediate call. For partial test-suite runs,
this can significantly impact run times.

Components: Vulkan

VK-GL_CTS Issue: 3460

Affects:
dEQP-VK.fragment_shading_rate.*

Change-Id: Ie2bb7a7690b2650006da888433397dae9a5ff7cf

Fix rounding issues with texture generation shader

VK-GL-CTS Issue: 3250

Affects:

dEQP-VK.graphicsfuzz.cov-global-loop-counter-texture-sample-loop-condition-set-array-element
dEQP-VK.graphicsfuzz.cov-multiple-fragcoord-conditions-false-never-return-sample-texture
dEQP-VK.graphicsfuzz.cov-nested-loops-sample-opposite-corners

Components: Vulkan
Change-Id: Ifba30c4dbf17f51c790ce66819da3785049bec6b

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: I064e8e9c25536c26df4b575da647443cc753e542

Re-enable GCC warning for Amber

The Amber version used by vulkan-cts-1.2.7 and later includes aa69a0
which should fix all uninitialized variable warnings.

Components: Framework, Vulkan

VK-GL-CTS issue: 3361

Change-Id: I95bab777232216dad1c1c02bd57f21c8bd2a1528

Fix rounding issues with texture generation shader

VK-GL-CTS Issue: 3250

Affects:

dEQP-VK.graphicsfuzz.cov-min-nested-loop-same-value-for-variables
dEQP-VK.graphicsfuzz.cov-sample-texture-hundred-iterations
dEQP-VK.graphicsfuzz.cov-sampler-as-function-argument
dEQP-VK.graphicsfuzz.cov-texel-double-negation
dEQP-VK.graphicsfuzz.cov-variable-copy-in-function-tex-sample
dEQP-VK.graphicsfuzz.spv-stable-sampler-loop-extra-instructions
dEQP-VK.graphicsfuzz.spv-stable-sampler-polar-simple-O-access-chain

Components: Vulkan
Change-Id: I1758bd50da844b2835c858ee5d78af4e7c4bd69a

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: Ib1155d0946f8c94eed1395127c6441c6a1a54174

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: I96cce7d1a37d9ab836f098aecf1f5895a9e0185d

Fix Set/WaitEvent2KHR dependency flags setup

Vulkan specifies that the dependencyFlags member of pDependencyInfo
must be 0 in case of vkCmdSetEvent2KHR and vkCmdWaitEvents2KHR.
The conformance tests use a non zero default flag.

Components: Vulkan

Affects:

dEQP-VK.synchronization2.basic.event.*
dEQP-VK.synchronization2.op.single_queue.event.*

VK-GL-CTS Issue: 3462

Change-Id: Ic67fb77bcddf198fc77bbc9b3a25979d3e43678b

Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5

Change-Id: Ic3ea84ef67493d7b596c4f4a9547b7841170a9a9

Fix compilation on Android

Function testSemaphoreImportSyncFdSignaled is not used and the build process
for Android fails because of it.
Fix just removes this unused function.

Components: Vulkan, AOSP

Affects:

None

VK-GL-CTS Issue: 3465

Change-Id: I82ad0e8df8196ac08d0003f30b7e624461239f84

Fix issues in load_store_op_none tests

This commit fixes several issues from the load_store_op_none tests:

- Checking depth/stencil image format support in a checkSupport() method
  instead of at iteration time.
- Validation error regarding mismatched attachment counts in dynamic
  rendering variants.
- Validation error regarding image layout in dynamic rendering variants.
- Wrong barriers after clearing color and depth/stencil images.
- Make the code not assume depth/stencil formats have both aspects,
  paving the way to testing more formats in the near future.

Affected tests:
dEQP-VK.*.load_store_op_none.*

Components: Vulkan
VK-GL-CTS issue: 3413

Change-Id: Iec237547f5762f299f1225eb4dda010e4256b367

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: Idb8b94c0b0e0052ff4bafffcdf473b1e9aa468e1

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: I078d372dbbcb1886aa13f37995bfcf50cd97413a

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: I60ad45aaeeb84782bde80829cb1e731fb8e76b01

Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5

Change-Id: Iff22de3540519a435bbb7874188bbcadec977b05

Fix bug in calcFloatDiff

bsign variable in function calcFloatDiff is wrongly declared. Fixing.

VK-GL-CTS issue: 3472

Affects:
dEQP-VK.api.image_clearing.core.clear_depth_stencil_image*
and all the tests from the vktApiImageClearingTests.cpp file.

Components: Vulkan
Signed-off-by: Nishit Shah <nishit.shah@samsung.com>
Change-Id: I654920c9de460eb04caaca988552f65da3aa2687

Relax precision for matrix precision tests

The Vulkan spec does not specify the precision for inverse and
determinant built in functions, however newly added tests effecitively
adds a precision requirement for them.

The precision required in these tests is too high for some valid
implementations of these builtin functions.

Relax the precision required in the tests to allow valid
implementations to pass.

Affects:
dEQP-VK.glsl.builtin.precision*.determinant.*
dEQP-VK.glsl.builtin.precision*.inverse.*

Compontents: vulkan

VK-GL-CTS issue: 3182

Change-Id: Ib4c5b9d04578732456b8f981bc82a2c86ad877fa

Skip the "two_draw" variants of zero_stride_with_offset tests

There's only one vertex in vertex array for singleVertex config.
Non-zero stride can cause out of bound memory access error since
vkd.cmdDraw draws 6(singleVertexDrawCount) vertices in singleVertex
config.

Affects:
dEQP-VK.pipeline.extended_dynamic_state.*.zero_stride_with_offset

Components: Vulkan

VK-GL-CTS issue: 3463

Change-Id: I77baa395c40f74f518bc9c7e8daf32c87686868a

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: I4e8734003d9e689045fd25601bc521a4fb119f0b

Merge vk-gl-cts/vulkan-cts-1.2.6 into vk-gl-cts/vulkan-cts-1.2.7

Change-Id: Iab61e615b3f7d5b108d925150255a73171e3d438

Merge vk-gl-cts/vulkan-cts-1.2.5 into vk-gl-cts/vulkan-cts-1.2.6

Change-Id: If56557dcb21ca8bfd90acbb2fffbd3f8ceed16f2

Re-enable Clang warning for Amber

The Amber version used by vulkan-cts-1.2.5 and later includes 4d0115c
which should fix all suggest-destructor-override warnings.

Components: Framework, Vulkan

VK-GL-CTS issue: 3361

Change-Id: I36437c797c067521e1444d20cfece33babfe435a

GCC 11 build fixes

Some files used std::numeric_limits without including <limits> and
failed to compile.

VK-GL-CTS issue: 2842

Change-Id: I21e37ce38db08c7c7041f4628ac852588a788852

Merge vk-gl-cts/vulkan-cts-1.2.4 into vk-gl-cts/vulkan-cts-1.2.5

Change-Id: I790e54266657e91c84b9e40f6eb7385914d569c6

Decrease required subgroup size

dEQP-VK.spirv_assembly.instruction.spirv1p4.entrypoint.comp_workgroup_entry_point
test assumed that the subgroupSize is >= 8. The test failed if the
supported subgroupSize is less than 8.

This CL lowers the required subgroup size to 4.

Affects:
dEQP-VK.spirv_assembly.instruction.spirv1p4.entrypoint.comp_workgroup_entry_point

Components: Vulkan

VK-GL-CTS issue: 2698

Change-Id: I9931dab7a57c0b6b73530e72bca0c0d9cc90f87d
(cherry picked from commit 8f9b9137470ed5b0cd0b38edd9e9c28ccf565731)

Fix for potentially uninitialized variables

This change adds a constructor for a struct which sets the members that
do not have a constructor of their own to known values.

This potentially affects anything that uses the reference rasterizer,
but should not change any test results.

Affects:
*

Components: Framework
VK-GL-CTS issue: 3434

Change-Id: Ie874b4baa4d2ff5a8aa66f9ea2c3cd11b37edb64

Merge vk-gl-cts/vulkan-cts-1.2.3 into vk-gl-cts/vulkan-cts-1.2.4

Change-Id: I3080cbdc60602f0faea714502bbc5461cac257e0

Fix NULL/False to std::string conversion errors

Newer versions of GCC reject std::string(DE_NULL). Previously this
didn't cause any errors or warnings when compiling, but should produce
an error at runtime. This change replaces invalid values with empty
string literals and adds deFatalStr for cases that potentially
relied on the string constructor throwing an exception.

Components: Framework, OpenGL, Vulkan
VK-GL-CTS issue: 3419

Change-Id: I68ec7d1447d1e96f5d7841e0684687cdd294fe39

Require geometry shaders for adjacency topologies

Clipping tests would use adjacency topologies with drivers that didn't
support geometry shaders.

Components: Vulkan

Affects: dEQP-VK.clipping.clip_volume.depth_clip.*
VK-GL-CTS Issue: 2906

Change-Id: I7a41e62dcbcdcab46ce4f8b04bfe322c688fbcac

Remove redundant load/store op none tests

Load/store op none tests had originally a hard coded
depth stencil format, but was later changed to iterate
on multiple formats. The depth tests were accidentally
left there (as a duplicate) outside a format loop, and
the depth format parameter was uninitialized in these
tests.

This change removes these old tests and also fixes
the dynamic rendering tests to use correct depth
stencil formats.

VK-GL-CTS Issue: 3426

Affects:

dEQP-VK.*.suballocation.load_store_op_none.*

Components: Vulkan
Change-Id: I5c1e0a6d95bb1dbc88364f6e99f84e47619c285d

Fix vktFragmentShadingRateBasic tests

Many graphics pipelines are used in dynamic render pass which uses
fragment shading attachment, but these pipelines don't have
VK_PIPELINE_CREATE_RENDERING_FRAGMENT_SHADING_RATE_ATTACHMENT_BIT_KHR
flags when they are created. Adding those flags.

Affects: dEQP-VK.fragment_shading_rate.dynamic_rendering.*

Components: Vulkan

VK-GL-CTS issue: 3440

Change-Id: I3dd0aae9d6e9f2fc0bb5266f4f4ba442ebc0317b

Merge vk-gl-cts/vulkan-cts-1.2.7 into vk-gl-cts/vulkan-cts-1.2.8

Change-Id: Ic1cf8fed8e0713c216ff31212334cb53acbbf72d

Fix getting capture address in RT capture replay tests

Use vkGetBufferOpaqueCaptureAddress to retreive address
used in opaqueCaptureAddress

Components: Vulkan

VK-GL-CTS Issue: 3420

Affects: dEQP-VK.ray_tracing_pipeline.capture_replay.shader_binding_tables
Change-Id: Ifaa2fe92d6517fa75a29ad21c06365efd7857de4

Remove time measurement on deferred host build tests

The time measurement is wrongly implemented in this test
- it measures also thread creation, allocations and VK setup time, additionally
to AS build.  It is not easily fixable.

Moreover, the spec says that offloading the work to background threads
is 'desirable', however it never says it is faster.

VK-GL-CTS Issue: 3435

Components: vulkan

Affects:
dEQP-VK.ray_tracing_pipeline.acceleration_structures.*
dEQP-VK.ray_query.acceleration_structures.*

Change-Id: I41d836e2e454b9fb9bb46109edddd4eda19370e9