Relax protected heap allocation tests - continuation

Some implementations have limitations on protected heap
allocations, and these limitations cannot be queried
using the Vulkan API. This change allows an out-of-memory
to happen after a certain number of allocations to
compensate that.

Fix protected memory allocation to be done on small
allocation chunks and ignore allocation failures
if there is at least 80 of concurrent allocations
at a time.

VK-GL-CTS issue: 3257

Affects:
dEQP-VK.memory.allocation.random.*

Components: Vulkan
Change-Id: I32ed020195ff468de468bbd8527dec1d7f0db183

GCC 11 build fixes

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

Some dynamic casts and calls to AndroidHardwareBufferExternalApi::
getInstance() produced nonnull warnings. Suppress these by
wrapping them in if statements.

This commit also removes implicit-fallthrough from the list of ignored
GCC warnings in check_build_sanity.py. All such warnings in the CTS
should be fixed by now and recent versions of Clang also support
this warning.

VK-GL-CTS issue: 2842

Change-Id: I5bd290f9c312a30c2fcb223307c7b5c785a0fa92

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

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

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

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

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

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

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

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)

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

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

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

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: Id485bc8ae929179e7786039966a233dd6fa57cdc

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)

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>

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>

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: Icdbb121ec7a828457ff9df34efe018fbf845e3f9

Accept NaN as valid pass value in shader precision atan2 tests

Spec mentions that implementation may assume arguments are not
Inf if the entrypoint is not declared with SignedZeroInfNanPreserve
and can perform optimizations.

Components: Vulkan

VK-GL-CTS Issue: 3073

Affected tests:
dEQP-VK.spirv_assembly.instruction.compute.float16.*.atan2
dEQP-VK.spirv_assembly.instruction.graphics.float16.*.atan2*
dEQP-VK.glsl.builtin.precision.atan2.*

Change-Id: I6e1fbc14a5654422cf88e1453fae0774bfce0bb6

Fix AHB external memory test mip levels

AHB external memory tests were testing mipmaps using a
queried maximum number of supported mipmap levels, but
at the same time using an image size of 64x64.
The number of mip levels need to match the number for a
full mip chain, in this case 7. Now a correct number of
mip levels is used, which also gets rid of a validation
error.

VK-GL-CTS Issue: 3112

Affects:

dEQP-VK.api.external.memory.android_hardware_buffer.image_formats.*

Components: Vulkan
Change-Id: I1c6d09f93c06ac1c52cea283fd86359ccde01e4f

Remove invalid SPIR-V decorations

Some SPIR-V asm tests used invalid Binding and DescriptorSet decorations.
This CL removes those decorations.

Affects:

dEQP-VK.spirv_assembly.instruction.compute.physical_pointers.complex_types_compute.*
dEQP-VK.spirv_assembly.instruction.compute.pointer_parameter.buffer_memory
dEQP-VK.spirv_assembly.instruction.graphics.pointer_parameter.buffer_memory*
dEQP-VK.spirv_assembly.instruction.graphics.16bit_storage.input_output_float_16_to_16x2.scalar
dEQP-VK.spirv_assembly.instruction.graphics.16bit_storage.input_output_float_16_to_16x2.vec2
dEQP-VK.spirv_assembly.instruction.graphics.16bit_storage.input_output_int_16_to_16x2.scalar_int
dEQP-VK.spirv_assembly.instruction.graphics.16bit_storage.input_output_int_16_to_16x2.scalar_uint
dEQP-VK.spirv_assembly.instruction.graphics.16bit_storage.input_output_int_16_to_16x2.ivec2
dEQP-VK.spirv_assembly.instruction.graphics.16bit_storage.input_output_int_16_to_16x2.uvec2

Components: Vulkan

VK-GL-CTS issue: 3063

Change-Id: Icfee33f9ce041bf60a19fba55aef9fdab0747148

Merge remote-tracking branch 'origin/vulkan-cts-1.2.2' into vulkan-cts-1.2.3

Change-Id: I3c5b289149d9f2601b5d5ab8986d41b3f33f8bab

Correct Vulkan multiview 'tesellation' test names

Tests are renamed from 'tesellation' to 'tessellation' but are otherwise
unchanged.

Affects:

dEQP-VK.multiview*tessellation*

Components: Vulkan
Change-Id: I1686bb6fb8a0e85e201853602f840e03e847169d

Fix protected memory ssbo pipeline barrier use

Incorrect stage flags are passed to vkCmdPipelineBarrier in the copy
buffer operation in the dEQP-VK.protected_memory.ssbo.ssbo_read.*

These tests were introduced in origin/vulkan-cts-1.1.0

Components: Vulkan

Affects:
dEQP-VK.protected_memory.ssbo.*

VK-GL-CTS Issue: 3041

Change-Id: Iccdb96fee30e32bf26dec1e26d65ea8d5600c384

Fix error in build_apk.py

A typo caused build_apk.py to fail if required Android build
tools were missing.

Components: framework

Change-Id: Ia55b95a9057a37152e5a4c91773b1eaa3abc0156

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: Iab13500b9aaa9b6da0e99ea9d93fe9cabc66da33

Relax protected heap allocation tests

Some implementations have limitations on protected heap
allocations, and these limitations cannot be queried
using the Vulkan API. This change allows an out-of-memory
to happen after a certain number of allocations to
compensate that.

VK-GL-CTS Issue: 3038

Affects:

dEQP-VK.memory.*allocation.*

Components: Vulkan
Change-Id: I3a34637c8c15eb8670613dce9b6e394a622a4c71

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: If1bb9f63d56815c3ced2ac72d17d4e55ec090f39

Fix early destruction of VkPipelineLayout

Components: Vulkan
VK-GL-CTS issue: 3004

Affected:

    dEQP-VK.sparse_resources.shader_intrinsics.*

Change-Id: Id5508fac89961c5bb934005257d2caa26d32f2dd
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>

Fix validation errors in AHB tests

AHB protected memory tests were also removed as the test
tried to use protected memory without creating a device
and queue that would support protected operations.

VK-GL-CTS Issue: 2508

Affects:

dEQP-VK.api.external.memory.android_hardware_buffer.*

Components: Vulkan
Change-Id: I307e7450d06d20b6497f2a529e5724ae0538e5a9

Clean ExternalMemoryUtil

This change separates choosing of memory type from
allocateExportableMemory utility function. This makes
use of this functions more clear.
Previous implementation needed to be done ASAP and
involved passing of non-const reference to
allocateExportableMemory which was then filled.
This workaround made code unnecessarily hard to follow.

Components: Vulkan

VK-GL-CTS issue: 587

Affects:
dEQP-VK.api.external.memory.*
dEQP-VK.synchronization.signal_order.*
dEQP-VK.synchronization.cross_instance.*

Change-Id: I781388db390f13c7c03785dd0ae34ae0128ec31c

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: Ic61d3218b466b8fd2c50127867a0cfdb77c07637

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: I7c0abe6efdeab69b5928cc0e67a6365f15b1e5c5

Call computeFixedPointError only for fixed point case

Calling computeFixedPointError for integer case (numAccurateBits=32)
is incorrect and causes crash.
Added assert to make sure numAccurateBits is in legal range.

Affected tests: dEQP-VK.sparse_resources.image_sparse_residency.*

Components: Vulkan

VK-GL-CTS issue: 2994

Change-Id: Ib357cbbbda0f7980ec2d7115053cabc6825a18d3

Do not exceed heap size when allocating memory

It is invalid to make larger allocations than heap size.

Components: Vulkan

VK-GL-CTS Issue: 2927

Change-Id: I738356254d95a152a2bb89d711eb6b3f047b405c
Affects: dEQP-VK.memory.mapping

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: I6a05c3d94da3f1f6397b029997d6d43db2439548

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: Ib9e1ec8639e95a5bda7c361e1db102b5f4f94194

Fixes ycbcr copy tests using NaNs

Removes NaNs from the source images

Components: Vulkan
VK-GL-CTS issue: 2772

Affects:
dEQP-VK.ycbcr.copy.*

Change-Id: I52d279633e081ecce0ef93681c4d199296de6107

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: I9a085c6867ee119951a839b0d5eecb7fcf351029

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: I575ea750c7a89ec9f8306537a4bc3f7fbb64a619

Fixes sparse image padding tests

Tests were attempting to verify the
values inside the padding X bits of
formats like 10x6, where spec says
they are undefined.

Components: Vulkan
VK-GL-CTS Issue: 2916

Affects:
dEQP-VK.sparse_resources.image_sparse_binding.2d*

Change-Id: Ic785fba111ecebb5a030673642b78872c404d925

Fixes fragment density map tests

Multiview extension struct being passed to
render pass 2 is not valid. Need to use
viewMask in the render pass 2 structs instead.

Components: Vulkan
VK-GL-CTS Issue: 2899

Affects:
dEQP-VK.renderpass2.fragment_density_map.*

Change-Id: Ieedaacae4fe7ba6d3c222610d96f1785fa01719a

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: Ib3ccf6adeedb2a55ba8a96bfabb75273a1c664ea

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: I90603cf34d44bb7c6f10e81f7b594834391da05e

Android: test list change dEQP-VK.glsl.builtin.*

Move the following tests from the Android 2019 test list to the 2020
test list.

dEQP-VK.glsl.builtin.precision.asin.highp.scalar
dEQP-VK.glsl.builtin.precision.asin.highp.vec2
dEQP-VK.glsl.builtin.precision.asin.highp.vec3
dEQP-VK.glsl.builtin.precision.asin.highp.vec4

Components: AOSP

Change-Id: I6586b0cfb39be6d43278cf61378563b3f894d641

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: I452f719e084e6decc02066064c6f0397e4b915e6

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: I2947e16b40c26feb3c95922e18e36e36acbc43b9

Merge vk-gl-cts/vulkan-cts-1.1.6 into vk-gl-cts/vulkan-cts-1.2.1

Change-Id: Id775a0b8cb86bd5929b61c49fbfac4825421c7a7

Remove in-repo verify_submission scripts

These scripts are now replaced by the CTS Tools repo:
https://github.com/KhronosGroup/VK-GL-CTS-Tools

Components: Vulkan, OpenGL

Change-Id: If0f47b4186ed10f9946852e8d37956d7328625a8

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: I6bad956acaf784565692e1ea539a219064deed57

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: I64b979022925b98680602154ad32b13080a04871

Merge vk-gl-cts/vulkan-cts-1.1.6 into vk-gl-cts/vulkan-cts-1.2.1

Change-Id: If6077d356659193696b2e41c318b5bda2f5ca1d1

Update Vulkan/GL Readme

Update the Readme files to point to the new
submission verifcation script

Components: Vulkan, OpenGL

Change-Id: I6fb35a383fcc49e543ae67822be22d744c5fcd9c
(cherry picked from commit 412f5a3363eb228ca1a63be337de172a2e622571)

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: I15d9b2b0dbaab175d891a824c190198bbcda1b20

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: Iccb63bb8842b2d3b8ae22b78853a0534500aebdf

Adds local multi device present mode support check

Test calls vkGetPhysicalDevicePresentRectanglesKHR even if
driver does not support
VK_DEVICE_GROUP_PRESENT_MODE_LOCAL_MULTI_DEVICE_BIT_KHR

Components: Vulkan
VK-GL-CTS Issue: 2828

Affects:
dEQP-VK.wsi.wayland.surface.query_devgroup_present_modes

Change-Id: I91ac8d2820318d1553e28fe71a4d2488a0fead5d

Fixes cubic filtering rgb9e5 tests

Test was causing overflows

Components: Framework
VK-GL-CTS Issue: 2771

Affects:
*

Change-Id: I3367d18dab3207442d93adb757db24c16543f6d8

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: I3e2556116ad51d257d5605321813461dc3096df1

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: Ic40e8dc5399b4abedbc4663d67431e5dee093d29

Remove renderingComplete test

There is no guarantee that two consecutive frames with
no frame dependencies will not complete out of order.

Components: EGL
Affects: dEQP-EGL.functional.get_frame_timestamps
Test: atest CtsDeqpTestCases
Bug: https://issuetracker.google.com/174941611

Change-Id: Ia70b92be27e301e94e936d9465a6e9dc9ec38928

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: I866350c03d9b27e87b39653aef4415f326e2510a

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: Id4b318d832da2ccfbc5a72523454eb39a23d2a2c

Account for aliasing exceptions in codegen

Fixes a problem where functions for which aliasing exceptions are in
place (at present this is just glRenderbufferStorageMultisampleEXT)
were not being treated properly in code generation -- their aliased
names were being erroneously used.

Component: Framework

Affects: dEQP-GLES*multisample*

VK-GL-CTS issue: 2291

Change-Id: I5e9caf8828b192d15bb3c91f53a9d09aa9442129

Fix heap read out of bounds in RenderPass tests.

Pass the correct size of the buffer to uploadBufferData.

Affected tests:
dEQP-VK.renderpass*.attachment_allocation.*
dEQP-VK.renderpass*.suballocation.*

Components: Vulkan
VK-GL-CTS issue: 2708

Change-Id: Ib498be09f1c8437e8080c5b5f3a12d8b7a25a262

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: Ib199c6c60ed14c4549a90ad19dcb20a7a5ad1a27

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: I2321a93eaa56abbe87c87680aedd8ceb76b645e1

Merge vk-gl-cts/vulkan-cts-1.2.0 into vk-gl-cts/vulkan-cts-1.2.1

Change-Id: If26f7fae4e13aff58e233680c755db99879644b3

Merge vk-gl-cts/vulkan-cts-1.1.6 into vk-gl-cts/vulkan-cts-1.2.1

Change-Id: I3ed8d5f4de16750c664e4c8908b8b971f48f7593

Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6

Change-Id: I004160d9169fcb42a928c10d76b5355dd671e5a8

Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.2.0

Change-Id: Ie1ff8e21a9cffae5e0616204ff3416e490ca0637

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: I71fc9c2f2a1132010379b79b0ca2445f57c5055c

Remove cov-fold-div-uniform-by-zero

The test relies on undefined behavior

Components: Vulkan

VK-GL-CTS issue: 2691

Affects: dEQP-VK.graphicsfuzz.cov-fold-div-uniform-by-zero

Change-Id: Id5703afdf8aeb544b601184b91463a5cf0a7c532

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: Id9f7774598b101e0f88ce933f347aadffcd7c73d

Remove travis.yml

Components: Framework

Change-Id: I14bcef09f4c198dc14c56c304cbc1199b4110ce2

Remove cov-fold-div-variable-by-zero

The test relies on undefined behavior

Components: Vulkan

VK-GL-CTS issue: 2691

Affects: dEQP-VK.graphicsfuzz.cov-fold-div-variable-by-zero

Change-Id: I56f7e34bb237184b53c60b8074d19e16c400fff7

Fix slow run time of OpPhi wide test

dEQP-VK.spirv_assembly.instruction.compute.opphi.wide was using 1024
arguments for OpPhi SpirV instruction. This was causing very slow
run/compile time on some devices and therefore failing the test due to
timeout.

This CL halves the number of arguments given to the OpPhi instruction to
avoid timeouts.

Affects:

dEQP-VK.spirv_assembly.instruction.compute.opphi.wide

VK-GL-CTS Issue: 2681

Components: Vulkan
Change-Id: Ice41e983b8da866916e07c1b52d7ccfd65001077

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: Ie115d2e00cf468c6403fae07c9f77b466c3c4380

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: I9855d3d8134ac73ea00b95adde9a66a37c0a03b3

Fixes descriptor indexing tests cache issue

Tests were not invalidating the results
buffer data before verifying.

Components: Vulkan
VK-GL-CTS Issue: 2682

Affects:
dEQP-VK.descriptor_indexing.*

Change-Id: Ia2afa6697d12fd3f4ebc8ef29ae01d166f290a7f

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: I66f05ed7be1361806de4ea2d8e17ea6472bb93cb

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: I3966e7e1ab45fa815f85def0edaab713e6e16f80

Loosen basic subgroup elect test vertex checks

This commit takes into account any shader in the vertex pipeline may run
multiple times per vertex when checking results in the basic subgroup
elect test.

In addition, it contains a few indentation fixes as well as fixes for
several validation errors in the whole test group.

Affected tests:
dEQP-VK.subgroups.basic.*

Components: Vulkan
VK-GL-CTS issue: 2162

Change-Id: Ib0aaf929bca5ce4939a3e2fd98a69a0812f1aa0b

Add tests for VK_EXT_subgroup_size_control

For compute and graphics pipelines:

* Check the required subgroup size set by the tests is honored by the
  driver (including for full subgroups and allow varying cases for
  computer shaders).
* Check that enabling allow varying subgroup size reports a subgroup
  size that falls within the limits of the driver.
* For compute pipeline, check that require full subgroups works as
  expected.

On top of that, added requiredSubgroupSize case for subgroup's compute
shader tests that check all possible required subgroup sizes announced
by the driver.

Also there is a test that checks subgroup size control limits sanity.

Affected tests:

  dEQP-VK.subgroups.*

Components: Vulkan
VK-GL-CTS issue: 1864

Change-Id: Ia48deb9318420889754bd97545045f85b3c6fed7

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: Idb3171db18746b59d7c91297bd6ab3c00be455ae

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: I2cdeeb0ab92e4533ee38ef0d2c39a2845b03379a

Fixes texture filter cubic mipmap tests

Tests using wrong miplevel to verify

Components: Vulkan
VK-GL-CTS Issue: 2662

Affects:
dEQP-VK.texture.filtering.*.cubic

Change-Id: I65ca120bdcaf460ce4addf15bc5ed15bbc3bc7b4

Fix descriptorCount for YCbCr-related tests

Certain YCbCr tests did not behave consistently with the Vulkan spec
when creating descriptor pools.
Tests now properly query combinedImageSamplerDescriptorCount to
correctly set pool size in the event of multiplane images.

Affects:

dEQP-VK.ycbcr.filtering.*

Components: Vulkan

VK-GL-CTS Issue: 2386

Change-Id: Idd700bee41eec7546f903326c31fbc928c7e7ecb

Fixes sparse resource SNORM tests

Tests verifies value out of range

Components: Vulkan
VK-GL-CTS Issue: 2598

Affects:
dEQP-VK.sparse_resources.image_sparse_binding.*snorm*
dEQP-VK.sparse_resources.mipmap_sparse_residency.*snorm*

Change-Id: Id11addbc60c44ca7f20f86e11dbc7c833bbe7da7

Merge vk-gl-cts/vulkan-cts-1.2.2 into vk-gl-cts/vulkan-cts-1.2.3

Change-Id: Iaa3c6741b9d7ad77eda048ec234c59e869263182

Merge vk-gl-cts/vulkan-cts-1.2.1 into vk-gl-cts/vulkan-cts-1.2.2

Change-Id: I0ce2eae2a8a98a50d25e9e916dd0dde8e90e3288

Merge vk-gl-cts/vulkan-cts-1.2.0 into vk-gl-cts/vulkan-cts-1.2.1

Change-Id: Ie6ac4a6779ee2dd624590df63f152f661c089223

Merge vk-gl-cts/vulkan-cts-1.1.6 into vk-gl-cts/vulkan-cts-1.2.1

Change-Id: I8f4da52990ac1b7ae96414375e9bea652ddbc7d8

Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6

Change-Id: Ie294402abe02966f55089430d9557b5f0bc80c5f

Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.2.0

Change-Id: Icda35a4b6ca926511422ec96c4675b9012973ef1

Permit not flushing operands for FP16 DenormFlushToZero add, dot, vmulm

Where `d` is a denormal value such that `2 * d` is not denormal:

- The F16 OpFAdd tests assumed (d + d) is zero when DenormFlushToZero is
  used. This is not true if the operands (d) are not flushed to zero by
  the implementation, which is allowed by the Vulkan specification.
- The F16 OpDot test assumed dot(vec2(d, d), vec2(1, 1)) is zero. Since
  that is equivalent to (d + d), it has the same problem as OpFAdd.
- The F16 OpVectorTimesMatrix test made the same assumption as the OpDot
  test but for both columns of a 2-by-2 matrix, with the same problem.

These tests are changed to accept `2 * d` as a valid result alongside
zero, such that they will pass both on implementations which flush
operands to zero and on implementations which only flush results to
zero.

Modified tests:

    dEQP-VK.spirv_assembly.instruction.*.float_controls.fp16.input_args.denorm_*flush_to_zero*
    dEQP-VK.spirv_assembly.instruction.compute.float_controls.independence_settings.denorm_ind_*_fp16_flush_*

VK-GL-CTS Issue: 2644

Components: Vulkan

Change-Id: I0ef4a57915e3bdab100f55f52d21913f87af4927

Fix descriptorCount for YCbCr-related tests

Certain YCbCr tests did not behave consistently with the Vulkan spec
when creating descriptor pools.
Tests now properly query combinedImageSamplerDescriptorCount to
correctly set pool size in the event of multiplane images.

Affects:

dEQP-VK.protected_memory.interaction.ycbcr.*
dEQP-VK.ycbcr.format.*
dEQP-VK.ycbcr.plane_view.*
dEQP-VK.ycbcr.query.*

Components: Vulkan

VK-GL-CTS Issue: 2386

Change-Id: I8bb261a29c7e4ce82deced447767de915cee451a

Fix command buffer usage flags in dEQP-VK.query_pool.performance_query tests

The test is resubmitting command buffer few times in a loop, however it
is using beginCommandBuffer utility function, which sets the command buffer usage flag to
VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT by default. This commit changes that to
0 (no flags).

Moreover, VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT flag is removed, as it seems
The submits are serialized by a fence.

Additionally replaced raw vkCmdBeginCommandBuffr with beginCommandBuffer for clarity.

Components: Vulkan

VK-GL-CTS Issue: 2617

Affects:
dEQP-VK.query_pool.performance_query.query*
dEQP-VK.query_pool.performance_query.multiple_pools*

Change-Id: Iff25e4cc715582710f3ed23367a7fd8f9a23b29e

Amend Vulkan Conformance Submission Package Requirements

The current practice is that all bugfixes should be merged
into the CTS repo before being used in a submission.
This change promotes this practices to a requirement.

Compontents: Vulkan

Change-Id: I61eef5522aecf97a3d42c49777681eb671affe87

Splits long shader in memory model tests

Components: Vulkan
VK-GL-CTS Issue: 2625

Affects:
dEQP-VK.memory_model.*

Change-Id: I8478e268fa5988b015acd2dd70cd370bfc4c082c

Fix validation of DRM modifier compatibility with image properties

Vulkan spec states that in case of image creation with list of
DRM modifiers (VkImageDrmFormatModifierListCreateInfoEXT), each
of them must be compatible with the parameters in
VkImageCreateInfo and its pNext chain.

This change adds a mandatory check for modifier compatibility with
the specified image properties by querying
vkGetPhysicalDeviceImageFormatProperties in a place where it was
missing prior to the actual image creation.

Components: Vulkan

Affects: dEQP-VK.drm_format_modifiers.create_list_modifiers.*

VK-GL-CTS Issue: 2611

Change-Id: Iab5deff6c8cbe6d2c510224322f7be91c7874acf