Boris Zanin [Wed, 23 May 2018 09:25:05 +0000 (11:25 +0200)]
Fix UConvert validation error in mul_sdiv tests
With new spirv-tools the SPIR-V validator enforces type check
requirement in OpUConvert, making this test to fail.
Replace UConvert with SConvert for signed tests.
Update SPRV-Tools.
Update tests:
* dEQP-VK.spirv_assembly.type.*high_part_zero*
Components: Vulkan
VK-GL-CTS issue: 1186
Change-Id: I73172eda179148d56aa104eb35cdc5afc25a9953
Jeff Bolz [Sat, 17 Mar 2018 17:30:36 +0000 (12:30 -0500)]
New tests for VK_NV_shader_subgroup_partitioned
Forked from subgroup arithmetic tests, they are mostly the same but with different shaders
New Tests: dEQP-VK.subgroups.partitioned.*
Components: Vulkan
Change-Id: I74e4f9cf716cf0d7293e4259dedf116af750347a
Alexander Galazin [Mon, 28 May 2018 08:22:48 +0000 (10:22 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.1 into vk-gl-cts/master
Change-Id: Ica9b46f19f9c3b45d0553db9deec82c696f3033a
Alexander Galazin [Mon, 28 May 2018 08:15:07 +0000 (10:15 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.0 into vk-gl-cts/vulkan-cts-1.1.1
Change-Id: If0ab08c681fd90e94d179ded910a531dcc518861
Slawomir Cygan [Wed, 18 Apr 2018 13:56:37 +0000 (15:56 +0200)]
Fix VkPipelineLayout use after free
The test destroys pipeline layout, used to create the graphics pipeline,
before the command buffer using this pipeline is recorded.
This is invalid according to spec:
"A VkPipelineLayout object must not be destroyed while any
command buffer that uses it is in the recording state."
Components: Vulkan
VK-GL-CTS Issue: 1134
Affects:
dEQP-VK.draw.differing_interpolation.*
Change-Id: I2dd692cfb69afaf0b9d6e8d8a29b86b945b7eaba
Jari Komppa [Thu, 17 May 2018 10:27:30 +0000 (13:27 +0300)]
Made glsl loop tests a bit more robust
The way the loop test suite checks for a correct result depends on
rotating the input vector's components, giving only 4 possible results.
In other words, a subtly failing implementation still had a 1 in 4
chance of passing the test.
This change increments the vector on each iteration and then decrements
the vector after the loop by the number of expected loop iterations. The
correctness test remains unchanged, but if the number of iterations does
not match the expected, the resulting vector will have a wrong value.
Also added tests where loops are within different control flow blocks,
and tests that contain a do..while block which should never loop unless
the loop variable is off by one.
Affects:
dEQP-VK.glsl.loops.*
New tests:
dEQP-VK.glsl.loops.special.*.pre_fallthrough_vertex
dEQP-VK.glsl.loops.special.*.pre_fallthrough_fragment
dEQP-VK.glsl.loops.special.*.post_fallthrough_vertex
dEQP-VK.glsl.loops.special.*.post_fallthrough_fragment
dEQP-VK.glsl.loops.special.*.dowhile_trap_vertex
dEQP-VK.glsl.loops.special.*.dowhile_trap_fragment
dEQP-VK.glsl.loops.special.*.ifblock_vertex
dEQP-VK.glsl.loops.special.*.ifblock_fragment
dEQP-VK.glsl.loops.special.*.elseblock_vertex
dEQP-VK.glsl.loops.special.*.elseblock_fragment
Components: Vulkan
VK-GL-CTS issue: 1166
Change-Id: I266537d6b5816a9e574b5b99061a4413d6493f41
Mika Väinölä [Wed, 28 Feb 2018 14:51:05 +0000 (16:51 +0200)]
Fix validation errors in synchronization tests
Change a pipeline stage flag to one supported by
VK_ACCESS_COLOR_ATTACHMENT_READ_BIT and initialize
a VkExternalSemaphoreProperties struct.
Affects:
dEQP-VK.synchronization.cross_instance.*
dEQP-VK.synchronization.internally_synchronized_objects.
pipeline_cache_graphics
Components: Vulkan
VK-GL-CTS issue: 1046
Change-Id: I537933539b616c14942791accee689db40294dba
Alexander Galazin [Fri, 18 May 2018 09:51:14 +0000 (11:51 +0200)]
Fix multiview mutlisample tests
The multiview multisample tests used VK_SAMPLE_COUNT_2_BIT
without checking that it is supported.
The minimum requirement is to support 4-bit and 1-bit only.
Affects: dEQP-VK.multiview.multisample.*
Components: Vulkan
VK-GL-CTS issue: 1152
Change-Id: Id999cbb42a5961cca1f89c709c1f0bc8d6124645
Juan A. Suarez Romero [Wed, 16 May 2018 11:01:52 +0000 (11:01 +0000)]
GL_SRG8_ALPHA8 also supported in OpenGL ES 3.0+
dEQP-GLES2.functional.fbo.completeness.renderable.texture.color0.srgb8_alpha8
assumes that using GL_SRGB8_ALPHA8 in GLES2 should fail unless proper
extensions are supported.
But the OpenGL ES context created can be either 2.0 or an upper version,
and upper versions (3.0+) supports this format without extensions.
Thus, if OpenGL ES 3.0+ context is returned, this test should assume
that rendering is possible.
This is an extension of fix done in Change-ID
Ibb95736959ff9d5b0ff972ccf3983abc98795be3.
Affectd tests:
dEQP-GLES2.functional.fbo.completeness.renderable.texture.color0.srgb8_alpha8
Components: OpenGL
VK-GL-CTS issue: 1137
Change-Id: I733a8595b1eba7756787cde4a6bca2391dc345d7
Chris Forbes [Sat, 19 May 2018 01:50:47 +0000 (18:50 -0700)]
Fix barriers in dEQP-VK.api.fill_and_update_buffer.*
Change-Id: Icb7ac6167e358556178c864884e7c7c53afd455c
Affects: dEQP-VK.api.fill_and_update_buffer.*
Components: Vulkan
VK-GL-CTS: 1177
Chris Forbes [Fri, 18 May 2018 22:50:39 +0000 (15:50 -0700)]
Add missing self-dependency in color->input self-dep tests
Change-Id: Ibf400c91bbc59abb49795078804e35e50c59c7db
VK-GL-CTS: 1181
Components: Vulkan
Affects: dEQP-VK.renderpass.*
Chris Forbes [Fri, 18 May 2018 20:49:26 +0000 (13:49 -0700)]
Account for noncoherent atom size in renderpass memory flush
Change-Id: Ibffb14fcdbaba5e9fbeb8ba9d9aba9cc59db4b17
Components: Vulkan
Affects: dEQP-VK.renderpass.*
VK-GL-CTS: 1180
Chris Forbes [Fri, 18 May 2018 17:46:27 +0000 (10:46 -0700)]
Remove spurious TS/GS stage bits from renderpass test barriers
None of this code actually uses these stages, and we weren't checking
whether they were supported. This was invalid on devices that don't
support the TS/GS stages.
Change-Id: I06cd1df95f52060150e36d28bc0525f2d398aef0
VK-GL-CTS: 1180
Components: Vulkan
Affects: dEQP-VK.renderpass.*
Paavo Pessi [Thu, 19 Apr 2018 11:47:40 +0000 (14:47 +0300)]
Add SPIR-V tests for 16, 32 and 64-bit integers
New tests were added to test various SPIR-V bit operations,
comparators, constants and initializers with signed and unsigned
integers of different width.
In bit field and shift operations some of the parameters are not
required to have the same bit width as Base and Result. This kind of
operations are tested with all possible combinations of result type
and parameter bit width.
The result type of bit field operations is always a 32-bit integer.
New tests:
dEQP-VK.spirv_assembly.type.*.shift_right_logical*
dEQP-VK.spirv_assembly.type.*.shift_right_arithmetic*
dEQP-VK.spirv_assembly.type.*.shift_left_logical*
dEQP-VK.spirv_assembly.type.*.bitwise_or*
dEQP-VK.spirv_assembly.type.*.bitwise_xor*
dEQP-VK.spirv_assembly.type.*.bitwise_and*
dEQP-VK.spirv_assembly.type.*.not*
dEQP-VK.spirv_assembly.type.*.iequal*
dEQP-VK.spirv_assembly.type.*.inotequal*
dEQP-VK.spirv_assembly.type.*.ugreaterthan*
dEQP-VK.spirv_assembly.type.*.sgreaterthan*
dEQP-VK.spirv_assembly.type.*.ugreaterthanequal*
dEQP-VK.spirv_assembly.type.*.sgreaterthanequal*
dEQP-VK.spirv_assembly.type.*.ulessthan*
dEQP-VK.spirv_assembly.type.*.slessthan*
dEQP-VK.spirv_assembly.type.*.ulessthanequal*
dEQP-VK.spirv_assembly.type.*.slessthanequal*
dEQP-VK.spirv_assembly.type.*.bit_field_insert*
dEQP-VK.spirv_assembly.type.*.bit_field_s_extract*
dEQP-VK.spirv_assembly.type.*.bit_field_u_extract*
dEQP-VK.spirv_assembly.type.*.bit_reverse*
dEQP-VK.spirv_assembly.type.*.bit_count*
dEQP-VK.spirv_assembly.type.*.constant*
dEQP-VK.spirv_assembly.type.*.constant_composite*
dEQP-VK.spirv_assembly.type.*.constant_null*
dEQP-VK.spirv_assembly.type.*.variable_initializer*
dEQP-VK.spirv_assembly.type.*.spec_constant_initializer*
dEQP-VK.spirv_assembly.type.*.spec_constant_composite_initializer*
Components: Vulkan
VK-GL-CTS issue: 1096
Change-Id: Id1d6c5b741b52f5acc9c759e5cfcd8ac4825532d
Joan Bonet Orantos [Tue, 22 May 2018 08:09:05 +0000 (10:09 +0200)]
Fix dependencies in vulkan pipeline test
The srcStageMask field in the subpass dependency was not strict
enough. This can cause that the incoming subpasses, that try to copy
samples from this subpass, may try to read content that may not be
written yet.
Affects:
dEQP-VK.pipeline.multisample.min_sample_shading.min_*
Components: Vulkan
VK-GL-CTS issue: 1185
Change-Id: Ide5e3d2a709e8fe26d23ada1d20bf2eb27beffd8
Juan A. Suarez Romero [Mon, 14 May 2018 15:24:04 +0000 (15:24 +0000)]
Skip BC/DXT formats in Texture3D tests
These formats only supports 2D images.
Affects:
KHR-GLES2.texture_3d.compressed_texture.negative_compressed_tex_image
KHR-GLES2.texture_3d.compressed_texture.negative_compressed_tex_sub_image
Components: OpenGL
VK-GL-CTS issue: 1163
Change-Id: I806dccfdbb2969b51c0ee9a1010c04107154733a
Dzmitry Malyshau [Wed, 9 May 2018 18:15:26 +0000 (14:15 -0400)]
Fix VulkanPlatform compiling and linking on OSX
There where 2 problems with the code:
1. it broke upon the change to vk::Library interface,
since OSX is not yet CI tested
2. it didn't link against OpenGL library for Default target,
which appears to be required for anything on OSX
The latter was always an issue, but one we could work around locally.
I believe the proposed upstream patch is clean enough to avoid our hacks.
Change-Id: I9b570e3700031d92367a815baf5b32f751e248b3
Components: Framework
Dzmitry Malyshau [Wed, 9 May 2018 17:14:21 +0000 (13:14 -0400)]
Fix image blit mirrored comparisons
The old looping would not go through any iterations for mirrored offsets.
The looping has to be mirrored when the destination is mirrored, and the
source texture coordinate computation adjusted accordingly.
Affects:
dEQP-VK.api.copy_and_blit.core.blit_image.*.mirror*.*
Change-Id: I8feb1d8b90c8013b292865b0e3b20ef414b05920
Components: Vulkan
Graeme Leese [Wed, 25 Apr 2018 15:52:21 +0000 (16:52 +0100)]
Remove one copy of the ballot test source
The code for the ballot tests is the same for framebuffer and
non-framebuffer tests but it's all written out twice. Make both tests
use the same source.
Components: Vulkan
Affects: dEQP-VK.subgroups.ballot_broadcast.*
dEQP-VK.subgroups.ballot_other.*
Change-Id: I3f8ac60a3339fd974742d8a6a3eb540559e2a382
Dzmitry Malyshau [Mon, 7 May 2018 20:24:07 +0000 (16:24 -0400)]
Fix barrier and invalidation of the fill_buffer readback
The old code had a barrier waiting for a different memory region that is being
filled, and same goes for the memory invalidation call that follows.
It resulted in the range of size=0 being requeste for invalidation, which
doesn't include actual bytes filled by cmdFillBuffer that is being tested.
Affects:
dEQP-VK.api.fill_and_update_buffer.*
Change-Id: Id2d68dc65145b42ecb0f966c70d57db81e9acb4a
Components: Vulkan
Paavo Pessi [Thu, 3 May 2018 08:53:30 +0000 (11:53 +0300)]
Test copying data from/to tightly sized buffers
New tests were added to verify that the source and destination buffers
in vkCmdCopyBufferToImage and vkCmdCopyImageToBuffer respectively can
be just large enough to contain the accessed elements, specifically in
cases where the buffer size is not a multiple of buffer row length.
New tests:
dEQP-VK.api.copy_and_blit.*.image_to_buffer.tightly_sized_buffer
dEQP-VK.api.copy_and_blit.*.image_to_buffer.tightly_sized_buffer_offset
dEQP-VK.api.copy_and_blit.*.buffer_to_image.tightly_sized_buffer
dEQP-VK.api.copy_and_blit.*.buffer_to_image.tightly_sized_buffer_offset
Components: Vulkan
VK-GL-CTS issue: 1124
Change-Id: Ice8f542a37bc45f969e33eb148ad0e77f40f8dd8
Toni Merilehti [Thu, 3 May 2018 12:04:09 +0000 (15:04 +0300)]
Add tests for render targets with write masks
New tests verify correct rendering when using multiple render targets
with write masks.
New tests:
dEQP-VK.renderpass.suballocation.attachment_write_mask.*
dEQP-VK.renderpass.dedicated_allocation.attachment_write_mask.*
Components: Vulkan
VK-GL-CTS issue: 1082
VK-GL-CTS public issue: 91
Change-Id: I3b308bdcd83f7936c078ecedcf21bd6339dd1f28
Alexander Galazin [Fri, 18 May 2018 13:01:32 +0000 (15:01 +0200)]
Whitelist VK_KHR_get_display_properties2
Components: Vulkan
Affects: dEQP-VK.api.info.device.extensions
Change-Id: Ic064e283e164cad2c9785ffe6253a2c192419668
Jari Komppa [Fri, 4 May 2018 11:51:17 +0000 (14:51 +0300)]
Added more early-out hardware support checks
This patch adds more early-out hardware support checks to speed up the
conformance test especially on hardware that does not support some
features.
Affects:
dEQP-VK.api.pipeline.*
dEQP-VK.api.copy_and_blit.*
dEQP-VK.api.buffer_view.create.*
dEQP-VK.api.buffer.*
Components: Framework, Vulkan
VK-GL-CTS issue: 899
Change-Id: Ied10dfd4ce60eda4d1830a38f42b5f38f813d50f
Ari Suonpaa [Thu, 26 Apr 2018 09:27:44 +0000 (12:27 +0300)]
Add test for dynamically indexing output variable components
Added test where output variable component is dynamically selected
using input data. Graphics test framework was modified to support
code inside main function as the output variables are not visible
inside a test function. Fixed support for vec4 output variables in
the same framework. Previously it was resulting in a code that wasn't
accepted by the validator.
Affects:
dEQP-VK.spirv_assembly.*
Components: Vulkan
VK-GL-CTS issue: 1122
Change-Id: I78f149b0f709e1d11dd977cc432fec0e5f971bed
Khushal [Fri, 13 Apr 2018 23:53:42 +0000 (16:53 -0700)]
Add AHardwareBuffer support to external memory tests
Added tests:
* dEQP-VK.api.external.memory.android_hardware_buffer.*
Components: Vulkan
VK-GL-CTS issue: 738
cherry-picked from: https://gerrit.khronos.org/#/c/1455/
V4: fix broken handle lifetime (Thanks Tony)
Change-Id: I8c9911692c46b4dd622424b20bd1f9827fac6796
Alexander Galazin [Fri, 18 May 2018 11:12:32 +0000 (13:12 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.0 into vk-gl-cts/vulkan-cts-1.1.1
Change-Id: I4b35a41d666a707410f88f222c93b661f7d983a1
Dave Airlie [Mon, 30 Apr 2018 03:02:29 +0000 (13:02 +1000)]
Fix check for 16-bit extension for 16-bit conversion functions.
All the 16-bit conversion tests run even if the vulkan implementation
doesn't advertise the 16-bit storage feature. Request the 16-bit
storage feature for all the 16-bit tests.
Components: Vulkan
Affect:
dEQP-VK.spirv_assembly.instruction.*convert*16*
VK-GL-CTS issue: 1150
Change-Id: Icf58fdd0af5c757d44716b5cddd193f78827192f
Jesse Hall [Wed, 16 May 2018 23:56:54 +0000 (16:56 -0700)]
Limit protected swapchain images based on memory
A surface's maxImageCount isn't based on the swapchain image extents
or format, so it's valid for swapchain creation to fail for
out-of-memory reasons even when the image count is within the
supported range.
This is particularly problematic for protected swapchains, since many
implementations have a limit on the total amount of protected memory
that can be allocated which is much smaller than total device memory.
This change estimates how much memory is used by protected swapchain
images, compares that to the size of heaps that support protected
memory types, and doesn't try to create swapchains that would have
more images than would fit within that size.
Components: Vulkan
VK-GL-CTS Issue: 1167
Affects:
dEQP-VK.protected_memory.interaction.wsi.*.swapchain.create.min_image_count
Change-Id: I6cc3bd25f7f22de8424f9e3375ade35fac26a884
Piers Daniell [Fri, 4 May 2018 22:53:25 +0000 (16:53 -0600)]
Don't use invalid sType as canary value for arrays
Using an invalid sType in an array is a violation of Vulkan
valid usage and will fail the test if validation layers
are enabled.
This CL also ensures the property structs passed into the
get functions have their sType initialized correctly.
Affects: dEQP-VK.wsi.display.*
Components: Vulkan
VK-GL-CTS issue: 1155
Change-Id: I4b9b00248265b2e3c7d0b2b652e98d2f0f5a0902
Graeme Leese [Thu, 26 Apr 2018 14:06:12 +0000 (15:06 +0100)]
Clear image layer 0 in memory commitment test
Components: Vulkan
Affected: dEQP-VK.api.get_memory_commitment.memory_commitment
VK-GL-CTS issue: 1151
Change-Id: I95676f18f91470248ac19270778b897626a31f2c
Peter Quayle [Wed, 16 May 2018 13:30:38 +0000 (14:30 +0100)]
Fix improper use of vkCmdWaitEvents
dEQP-VK.synchronization.smoke.events was calling vkCmdWaitEvents inside
a render-pass, which is not allowed if the event is later set using
vkSetEvent.
This is the part of change 2421 which was lost when change 2362 was
merged to master.
Affects:
dEQP-VK.synchronization.smoke.events
Components: Vulkan
VK-GL-CTS issue: 1089
Change-Id: Iddb17805fbffc426ac370e058a217ee7daf08e6a
Jari Komppa [Tue, 15 May 2018 11:26:15 +0000 (14:26 +0300)]
Fixed some visual studio warnings
Added some casts to quiet some warnings given by visual studio, mainly
int-float conversions and signed-unsigned comparison mismatches.
Affects:
None
Components: OpenGL
VK-GL-CTS issue: 1164
Change-Id: I56735830f275ff50116b53051a9b9a6d333edf7f
Ari Suonpaa [Tue, 17 Apr 2018 11:13:24 +0000 (14:13 +0300)]
Add 16bit storage access chain tests
Add tests to index data inside an array of structures. Covers 16bit
floats, integers, and unsigned integers.
New tests:
dEQP-VK.spirv_assembly.instruction.compute.16bit_storage.uniform_16_to_32_chainaccess.*
Components: Vulkan
VK-GL-CTS issue: 893
Change-Id: Iec98775737f2c2885d97f346096dc95e6a652cef
Juan A. Suarez Romero [Mon, 16 Apr 2018 15:45:05 +0000 (15:45 +0000)]
Add tests for 16-bit OpSwitch
Extend OpSwitch testing to cover signed and unsigned 16-bit integer
cases.
Add tests:
dEQP-VK.spirv_assembly.type.scalar.u16.switch_*
dEQP-VK.spirv_assembly.type.scalar.i16.switch_*
Components: Vulkan
VK-GL-CTS issue: 995
Change-Id: Icb39d94fd6112be2e7a1d07d481c2098247babb1
Slawomir Cygan [Wed, 21 Mar 2018 16:32:50 +0000 (17:32 +0100)]
Add SPIR-V tests for return values of atomic functions
Most of atomic functions return values before applying
atomic operation - this change adds tests for that.
Return values are captured to separate buffer and
matched to the final result (and arguments).
Components: Vulkan
VK-GL-CTS Issue: 1092
New Tests:
dEQP-VK.spirv_assembly.instruction.compute.opatomic_return_values.*
Affects:
dEQP-VK.spirv_assembly.instruction.compute.opatomic*
Change-Id: I96390ed1b694f0aa2bd6d92868da46fef3d5e249
Caio Marcelo de Oliveira Filho [Wed, 7 Mar 2018 22:52:20 +0000 (14:52 -0800)]
Test using gl_Layer in Vertex and Tessellation shaders
Needs VK_EXT_shader_viewport_index_layer extension. Test works by
writing rectangles of different colors and positions in each layer,
then verifying the result.
New tests:
- dEQP-VK.draw.shader_layer.*
VK-GL-CTS issue: 1043
Components: Vulkan
Change-Id: I037b8b1283b467b755e94b54d43e68fa4531497e
Juan A. Suarez Romero [Thu, 19 Apr 2018 15:35:40 +0000 (15:35 +0000)]
GL_SRG8_ALPHA8 also supported in OpenGL ES 3.0+
dEQP-GLES2.functional.fbo.completeness.renderable.renderbuffer.color0.srgb8_alpha8
assumes that using GL_SRGB8_ALPHA8 format in RenderBuffer should fail
for OpenGL ES 2.0 because it is not supported in the specification,
unless either EXT_sRGB or EXT_sRGB_write_control are implemented, as
they add this format as supported.
But when creating the OpenGL context, it can be either 2.0 or an upper
version. If 3.0+ is returned, this version supports this format without
those extensions, and thus the test should pass.
So this change ensures that if OpenGL ES 3.0+ context is returned, then
the use of GL_SRGB8_ALPHA8 in RenderBuffer should work.
Affected tests:
dEQP-GLES2.functional.fbo.completeness.renderable.renderbuffer.color0.srgb8_alpha8
Components: OpenGL
VK-GL-CTS issue: 1137
Change-Id: Ibb95736959ff9d5b0ff972ccf3983abc98795be3
Ostrowski, Igor [Fri, 9 Feb 2018 10:56:15 +0000 (11:56 +0100)]
Add tests for VK_KHR_draw_indirect_count
This change add new tests for VK_KHR_draw_indirect_count
extension.
New Tests:
dEQP-VK.draw.indirect_draw.*.indirect_draw_count.*
dEQP-VK.draw.indirect_draw.*.indirect_draw_first_instance.*
dEQP-VK.draw.indirect_draw.*.indirect_draw_count_first_instance.*
dEQP-VK.draw.indirect_draw.*.indirect_draw_count_instanced.*
Affected tests:
dEQP-VK.api.info.device.extensions
Components: Vulkan, Framework
VK-GL-CTS issue: 1004
Change-Id: I3d53806eb1a45293a77b940d9c2fa540a6d83d50
(cherry picked from commit
7b47cd53c6656b866e7c3a3c0a590da85f3b2de1)
Tony Zlatinski [Thu, 3 May 2018 18:35:51 +0000 (13:35 -0500)]
Vulkan: Fix WIN32 NativeHandle importMemory leaks
importMemory(), incorrectly calls handle.disown()
that discards the reference that the handle originally
added via its copy constructor to the underlying memory
handle.
Affects:
dEQP-VK.api.external.memory.opaque_win32
Components: Vulkan
VK-GL-CTS issue: 1139
Change-Id: Idb5111b7efc2dfbf7d693d79b7ab57814329f33e
Alexander Galazin [Fri, 11 May 2018 14:45:49 +0000 (16:45 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.0 into vk-gl-cts/vulkan-cts-1.1.1
Change-Id: I8ddf54dafd3de02f14526e98aa135176a6d3f3a3
Alexander Galazin [Fri, 11 May 2018 13:28:20 +0000 (15:28 +0200)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/vulkan-cts-1.1.0
Change-Id: I06a95f88214707d9108a46f7374a5e8e0f9bdb59
Alexander Galazin [Fri, 11 May 2018 11:53:34 +0000 (07:53 -0400)]
Merge "Merge vk-gl-cts/aosp-deqp-dev into vk-gl-cts/master"
Slawomir Cygan [Wed, 18 Apr 2018 13:14:23 +0000 (15:14 +0200)]
Multiview: do not use inline renderpass with secondary command buffers
There is a regression introduced in 7b76766 alongside the new render
pass begin utility:
The new utility sets the render pass contents flag to 'INLINE' by
default. This was overlooked in some multiview tests cases, which use
secondary command buffers - the default should be overriden to
'SECONDARY_COMMAND_BUFFERS'.
Components: Vulkan
VK-GL-CTS Issue: 1133
Affects:
dEQP-VK.multiview.secondary_cmd_buffer_geometry
dEQP-VK.multiview.secondary_cmd_buffer
Change-Id: Ibe42ba4c3e6d2b43ed18ac3684ba480f9dc659a8
Mateusz Zych [Sat, 14 Apr 2018 00:35:25 +0000 (20:35 -0400)]
Fix pipeline barrier for clearing multiview attachment
Specify correct destination access mask (VK_ACCESS_TRANSFER_WRITE_BIT) and
destination pipeline stage (VK_PIPELINE_STAGE_TRANSFER_BIT) for
transfer operation clearing all Views (array layers) of multiview attachment
using vkCmdClearColorImage().
Components: Vulkan
VK-GL-CTS Issue: 1085
Affects:
dEQP-VK.multiview.*
Change-Id: Ic87cd1d5a4c7373c3678c96fe942a5dde46ea03f
Mateusz Zych [Sat, 14 Apr 2018 00:11:14 +0000 (20:11 -0400)]
Clear all Views of Framebuffer Attachment to reference background color
The VK_KHR_multiview spec defines that during implicit clear (LOAD_OP_CLEAR)
all Active Views of Framebuffer attachment are cleared.
ViewMask defines which Views are active during particular subpass,
therefore a View is considered active during RenderPass
if it was active in any subpass of that RenderPass.
Without explicit clear of inactive Views their content will be undefined,
making test results unreliable.
This change removes MultiViewReadbackTestInstance::beforeDraw() and
relies on MultiViewRenderTestInstance::beforeDraw()
to clear all Views (array layers) of framebuffer attachment
by calling vkCmdClearColorImage().
Components: Vulkan
VK-GL-CTS Issue: 1085
Affects:
dEQP-VK.multiview.readback*
Change-Id: I70d8b1c48cb157df2518b7d241a302c54d5842f3
Alexander Galazin [Fri, 11 May 2018 10:29:10 +0000 (12:29 +0200)]
Merge vk-gl-cts/aosp-deqp-dev into vk-gl-cts/master
Change-Id: Ie711cb881e2de39c6bc881ae0ad72e16301145a3
Alexander Galazin [Fri, 11 May 2018 09:09:40 +0000 (11:09 +0200)]
Freeze 3.2.4.x mustpass
Change-Id: I1ed8e5fc0860c1e01d2eb7ed66f1ba411c31eef1
Paavo Pessi [Fri, 20 Apr 2018 11:56:37 +0000 (14:56 +0300)]
Fix validation errors in image tests
Image tests using VK_IMAGE_CREATE_BLOCK_TEXEL_VIEW_COMPATIBLE_BIT_KHR
and VK_IMAGE_CREATE_EXTENDED_USAGE_BIT_KHR had a pipeline barrier with
an invalid pipeline stage flag. The flag was changed to a valid value.
Affects:
dEQP-VK.image.texel_view_compatible.compute.basic.image_load.*
dEQP-VK.image.texel_view_compatible.compute.basic.texel_fetch.*
dEQP-VK.image.texel_view_compatible.compute.basic.texture.*
dEQP-VK.image.texel_view_compatible.compute.basic.image_store.*
dEQP-VK.image.texel_view_compatible.compute.extended.image_load.*
dEQP-VK.image.texel_view_compatible.compute.extended.texel_fetch.*
dEQP-VK.image.texel_view_compatible.compute.extended.texture.*
dEQP-VK.image.texel_view_compatible.compute.extended.image_store.*
dEQP-VK.image.texel_view_compatible.graphic.basic.attachment_write.*
dEQP-VK.image.texel_view_compatible.graphic.basic.texture_write.*
dEQP-VK.image.texel_view_compatible.graphic.extended.attachment_write.*
dEQP-VK.image.texel_view_compatible.graphic.extended.texture_write.*
dEQP-VK.image.extended_usage_bit.attachment_write.*
dEQP-VK.image.extended_usage_bit.texture_write.*
Components: Vulkan
VK-GL-CTS issue: 1051
Change-Id: Ie76fd515f04462d7f6ec2c2474d0af65737ccd9c
Ari Suonpaa [Fri, 20 Apr 2018 09:24:31 +0000 (12:24 +0300)]
Randomize image layout in renderpass attachment allocation test
Renderpass attachment allocation tests were using subpass image
layout of VK_IMAGE_LAYOUT_GENERAL. Now the layout is randomized
whenever possible.
Affects:
dEQP-VK.renderpass.*.attachment_allocation.input_output.*
Components: Vulkan
VK-GL-CTS issue: 1136
Change-Id: If59d42c50a2809c661ad56f0226b230659bc471f
Graeme Leese [Thu, 8 Mar 2018 15:54:13 +0000 (15:54 +0000)]
Check ballot requirement in subgroup basic tests
If ballot operations are not supported then the tests should return
NotSupported.
Components: Vulkan
Affects: dEQP-VK.subgroups.basic.*_framebuffer
VK-GL-CTS issue: 1127
Change-Id: I2c72a0821799ec2b94c08d2791f9d013ef7d122c
Chris Forbes [Thu, 3 May 2018 17:11:46 +0000 (10:11 -0700)]
Merge remote-tracking branch 'aosp/master' into deqp-dev
Change-Id: I24e845d3dd6283d88fed3fb17db9a5379e620a99
Chris Forbes [Thu, 3 May 2018 17:11:30 +0000 (10:11 -0700)]
Merge commit '
1d8110d34037070f007347a41b8212088d44df1e' into deqp-dev
Change-Id: Ifbbc3a9c5ebc1c4bc343369c227c2254162533a1
Chris Forbes [Wed, 2 May 2018 23:31:10 +0000 (16:31 -0700)]
Move unspecified-precision tests from ES2/ES3 to ES3.1
Got significant pushback from vendors about enforcing these tests on
ES2/ES3 -- ends up being a vendor choice between breaking shipping
content or failing CTS.
Moved all of these tests to ES3.1 only as a compromise.
Change-Id: I5a827064472f73908b01fa85e0097bda21038136
Components: AOSP
android-build-prod (mdb) [Tue, 1 May 2018 17:27:58 +0000 (10:27 -0700)]
[automerger skipped] Merge "DO NOT MERGE: Merge remote-tracking branch 'khronos/vulkan-cts-1.0.2' into HEAD" into oreo-mr1-cts-dev
am:
ea8d8c6e51 -s ours
Change-Id: Iafc1890dde07fdf56154e1742ce2d59648bc02af
android-build-prod (mdb) [Tue, 1 May 2018 17:23:07 +0000 (17:23 +0000)]
Merge "DO NOT MERGE: Merge remote-tracking branch 'khronos/vulkan-cts-1.0.2' into HEAD" into oreo-mr1-cts-dev
Lloyd Pique [Mon, 30 Apr 2018 19:34:53 +0000 (12:34 -0700)]
Allow the window to be recreated in the middle of a test am:
30ef1bf331 am:
38611cff1b
am:
6eca3b9153
Change-Id: I9347f789b79bf1bbcb128aab72ff4e62a662537a
Lloyd Pique [Mon, 30 Apr 2018 19:30:50 +0000 (12:30 -0700)]
Allow the window to be recreated in the middle of a test am:
30ef1bf331
am:
38611cff1b
Change-Id: Ib6b7ff02fe67679472877a9cc5dff04f7d01a17a
Lloyd Pique [Mon, 30 Apr 2018 19:26:18 +0000 (12:26 -0700)]
Allow the window to be recreated in the middle of a test
am:
30ef1bf331
Change-Id: Ic2683d5ad4515c3d65339b5027e76b1574374548
android-build-prod (mdb) [Mon, 30 Apr 2018 17:36:57 +0000 (10:36 -0700)]
[automerger skipped] Merge "DO NOT MERGE: Merge remote-tracking branch 'khronos/vulkan-cts-1.0.2' into HEAD" into oreo-cts-dev am:
354706417a -s ours
am:
c78fc14ea5 -s ours
Change-Id: Iedf43f729df806befedad07e7f8ae5d7b4bddd92
android-build-prod (mdb) [Mon, 30 Apr 2018 17:32:26 +0000 (10:32 -0700)]
[automerger skipped] Merge "DO NOT MERGE: Merge remote-tracking branch 'khronos/vulkan-cts-1.0.2' into HEAD" into oreo-cts-dev
am:
354706417a -s ours
Change-Id: I37e6dad1ab2d2eb9bee4a529f16c77981afbb54b
android-build-prod (mdb) [Mon, 30 Apr 2018 17:30:06 +0000 (17:30 +0000)]
Merge "DO NOT MERGE: Merge remote-tracking branch 'khronos/vulkan-cts-1.0.2' into HEAD" into oreo-cts-dev
Chris Forbes [Tue, 27 Feb 2018 00:31:59 +0000 (16:31 -0800)]
DO NOT MERGE: Merge remote-tracking branch 'khronos/vulkan-cts-1.0.2' into HEAD
Change-Id: Id624bed0598e117aff161ee6ed92a135095e33d7
Bug: b/
73945443
Chris Forbes [Fri, 27 Apr 2018 21:50:03 +0000 (14:50 -0700)]
Merge remote-tracking branch 'khronos/master' into deqp-dev
Change-Id: I7ae8d3a5d6fd711f7dc63c09b0c5d2fde386bb65
Chris Forbes [Fri, 27 Apr 2018 21:34:56 +0000 (14:34 -0700)]
Merge remote-tracking branch 'aosp/master' into deqp-dev
Change-Id: I868b09888968cdc3e233f5eef87967faa18b9b8e
Liam Middlebrook [Thu, 26 Apr 2018 21:43:25 +0000 (14:43 -0700)]
Do not fail plane tests when there is no display
Instead, we should mark these tests as not supported. The Vulkan spec
states that:
Devices must support at least one plane on each display
If there are no displays connected, the driver is allowed to return 0
planes.
VK-GL-CTS issue: 1138
Change-Id: Ie7ddc35f4f11fbfa54f8c31cc3864f52adb1bb7f
Affects: dEQP-VK.wsi.display.get_display_plane_capabilities
Affects: dEQP-VK.wsi.display.create_display_plane_surface
Damien Leone [Mon, 30 Oct 2017 23:15:25 +0000 (16:15 -0700)]
Do not fail display tests when there is no display
Instead, we should mark these tests as not supported. It is valid per the
Vulkan spec to have the extension report 0 displays connected.
Affects: dEQP-VK.wsi.display.*
Component: Vulkan
VK-GL-CTS issue: 1105
Change-Id: Iaae928a862dcc3827094d898a5455ecfe3920936
Slawomir Cygan [Tue, 24 Apr 2018 15:50:19 +0000 (17:50 +0200)]
Do not test CLEAR loadOp with read-only depth stencil image layouts
It is invalid to use implicit clears with depth stencil images, which first
render pass usage is in read only depth/stencil image layout
Spec:
"
For any member of pAttachments with a loadOp equal to VK_ATTACHMENT_LOAD_OP_CLEAR,
the first use of that attachment must not specify a layout equal to
VK_IMAGE_LAYOUT_DEPTH_READ_ONLY_STENCIL_ATTACHMENT_OPTIMAL.
For any member of pAttachments with a stencilLoadOp equal to VK_ATTACHMENT_LOAD_OP_CLEAR,
the first use of that attachment must not specify a layout equal to
VK_IMAGE_LAYOUT_DEPTH_ATTACHMENT_STENCIL_READ_ONLY_OPTIMAL."
Component: Vulkan
Affects:
dEQP-VK.renderpass.suballocation.formats.d*.clear.*_read_only
VK-GL-CTS Issue: 1144
Change-Id: Iad78473fe4bdf87dac5aff5bcc1c2b888e790cb8
Jari Komppa [Tue, 24 Apr 2018 09:08:28 +0000 (12:08 +0300)]
Fix logging issues when shader cache is enabled
When shaders were loaded from cache, log output said that linking
failed. This change fixes this and outputs the cache string as the
shader source. The info log also now says that the shader was loaded
from cache.
Affects:
None
Components: Framework, Vulkan
VK-GL-CTS issue: 1142
Change-Id: Ia9e060572c80a35152838fda1d5fbfd0fb6051a8
Lloyd Pique [Wed, 25 Apr 2018 02:08:39 +0000 (19:08 -0700)]
Allow the window to be recreated in the middle of a test
On ARC++ devices, some of the dEQP CTS tests were occasionally failing on
devices running in tablet mode, due to the timing of a WM relayout that
happens shortly after the orientation change is realized.
We traced it back to the dEQP test harness always expecting to see a
call from the system to onNativeWindowResized() after a call to
oNNativeWindowCreated(). It happened that the relayout on ARC++ devices
was not causing a size change, so this message was never received.
The result was that the test paused itself, and never completed. The
watchdog thread that the test harness creates ended up killing and
failing the test.
This patch adjusts the logic so that the activity only waits to process
the window creation for the very first call to onNativeWindowResized(),
as when the activity is first created that call is indeed made (but note
not really guaranteed). A subsequent call to onNativeWindowCreated()
will unpause the test (setting the READY state), as well as finish
processing the window creation.
Bug:
74371892
Test: dEQP-GLES3.functional.prerequisite#clear_color
Change-Id: I3c429aea272cbfee3f150a32c12233c9f38a7936
Chris Forbes [Tue, 27 Feb 2018 00:31:59 +0000 (16:31 -0800)]
DO NOT MERGE: Merge remote-tracking branch 'khronos/vulkan-cts-1.0.2' into HEAD
Change-Id: Id624bed0598e117aff161ee6ed92a135095e33d7
Bug: b/
73945443
Alexander Galazin [Tue, 24 Apr 2018 13:58:35 +0000 (15:58 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.1 into vk-gl-cts/master
Change-Id: I50e88f5187b0d970928fc00629158ceebd4659b4
Alexander Galazin [Tue, 24 Apr 2018 13:58:31 +0000 (15:58 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.0 into vk-gl-cts/vulkan-cts-1.1.1
Change-Id: I64646658117df9f4dd0260d0e0065e3f093b93d8
Paavo Pessi [Fri, 23 Mar 2018 13:32:30 +0000 (15:32 +0200)]
Test OpTypeImage with different depth property values
SPIR-V specification states that whether or not depth comparisons are
actually done depends on the sampling opcode, not the depth property
of the image itself.
New tests were added to verify that all three depth property values of
OpTypeImage can be used with both color and depth images, and that
depth comparisons are actually performed regardless of the depth
property.
New tests:
dEQP-VK.spirv_assembly.instruction.graphics.image_sampler.depth_property.*
Components: Vulkan
VK-GL-CTS issue: 1072
VK-GL-CTS public issue: 88
Change-Id: I5e0dbfba991ff1d477f5ebb706ffe1db32a2349f
Paavo Pessi [Thu, 5 Apr 2018 12:18:58 +0000 (15:18 +0300)]
Test copying data between compressed and uncompressed images
New tests were added to verify that when copying data between
compressed and uncompressed formats the extent members represent the
texel dimensions of the source image and not the destination.
Helper functions for querying information about compressed formats were
moved from vktImageTestUtil to vkImageUtil.
New tests:
dEQP-VK.api.copy_and_blit.core.image_to_image.all_formats.color.*
Components: Vulkan, Framework
VK-GL-CTS issue: 1081
VK-GL-CTS public issue: 90
Change-Id: Ie639e5d96a478ec1d8ef5c4abc73aa80a8005219
Ari Suonpaa [Fri, 13 Apr 2018 05:31:18 +0000 (08:31 +0300)]
Add 16bit to 16bit StorageInputOutput16 tests
Added 16bit storage tests where both inputs and outputs are 16bit.
New tests:
dEQP-VK.spirv_assembly.instruction.graphics.16bit_storage.input_output_*_16_to_16.*
Components: Vulkan
VK-GL-CTS issue: 893
Change-Id: I19b2e8f025b41a5b0a5d7d4699c7f101210e3c05
Jarred Davies [Thu, 12 Apr 2018 12:57:14 +0000 (13:57 +0100)]
Remove 4K min size from dedicated image tests
The tests dEQP-VK.memory.mapping.dedicated_alloc.image.*
all create an image with a minimum size of 4Kx4K, so the
tests in this subgroup test the exact same image size
regardless of what size the test itself is supposed to be
testing.
This change removes the 4Kx4K minimum to enable
the test to actually test multiple sizes of images.
Affect:
dEQP-VK.memory.mapping.dedicated_alloc.image.*
Components: Vulkan
VK-GL-CTS issue: 1118
Change-Id: I2ba9fa671f20eeb9bfa8b14972c903ac20875425
Panagiotis Apostolou [Mon, 26 Mar 2018 11:21:26 +0000 (13:21 +0200)]
Missing pipeline barriers from fragment variants of dEQP-VK.subgroups.*
This commit adds a missing image pipeline barrier in the fragment variant
of the subgroups test group. After the end of the renderpass and before the
vkCmdCopyImageToBuffer, a pipeline barrier is required to delay the copy
command so that all the writes to the color attachment are complete.
Tests affected:
dEQP-VK.subgroups.*
Components: Vulkan
VK-GL-CTS issue: 1135
Change-Id: I82a5179f97cf4435f96dd378f123f37bc26bd082
Alexander Galazin [Sun, 22 Apr 2018 20:39:57 +0000 (22:39 +0200)]
Merge vk-gl-cts/opengl-es-cts-3.2.5 into vk-gl-cts/master
Change-Id: I2982ab13ab697578f4206db28a55074b4344e68d
Alexander Galazin [Sun, 22 Apr 2018 20:39:09 +0000 (22:39 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.1 into vk-gl-cts/master
Change-Id: I2545688323c866d15daf56e4bc405889a4a8978c
Alexander Galazin [Sun, 22 Apr 2018 20:39:05 +0000 (22:39 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.0 into vk-gl-cts/vulkan-cts-1.1.1
Change-Id: I7a172b0447b8e67d7384dada152f7eb489d03f79
Jari Komppa [Mon, 16 Apr 2018 12:19:38 +0000 (15:19 +0300)]
Skip shader compile on unsupported ycbcr tests
Added a mechanism to check whether a test is supported before building
the shaders for the test. This also skips many of the tests earlier than
before.
This change also uses said mechanism to speed up the ycbcr tests.
Affects:
dEQP-VK.ycbcr.*
Components: Framework, Vulkan
VK-GL-CTS issue: 1066
VK-GL-CTS issue: 899
Change-Id: I477668e07a611349314f3eaf7ed741f28fb3ad0c
Dzmitry Malyshau [Fri, 2 Mar 2018 19:12:15 +0000 (14:12 -0500)]
VulkanPlatform for OSX
The change modularizes the OSX platform framework by splitting the GL
platform implementation out (into tcu::OSXGLPlatform) and adding a
basic Vulkan platform implementation (in tcu::osx::VulkanPlatform).
It is currently missing the WSI integration, and the memory limits may
need to be revised in follow-ups. This change allows the CTS to be run
against a Vulkan portability library on MacOS.
Components: Framework
Change-Id: I67eb9af5b97de701b4188aedcaca7485276d9001
Alexander Galazin [Thu, 19 Apr 2018 11:02:12 +0000 (13:02 +0200)]
Merge vk-gl-cts/opengl-cts-4.6.0 into vk-gl-cts/master
Change-Id: I7cfcc72895a1962b37815577c62fe8f0638728ee
Ari Suonpaa [Tue, 10 Apr 2018 05:31:16 +0000 (08:31 +0300)]
Use utility function for VkViewport and VkRect2D creation
Use makeViewport and makeRect2D whenever possible to make code a
bit more compact.
Affects:
dEQP-VK.api.*
dEQP-VK.draw.*
dEQP-VK.dynamic_state.*
dEQP-VK.geometry.*
dEQP-VK.image.*
dEQP-VK.multiview.*
dEQP-VK.pipeline.spec_constant.*
dEQP-VK.protected_memory.*
dEQP-VK.query_pool.*
dEQP-VK.subgroups.*
dEQP-VK.synchronization.*
dEQP-VK.tessellation.*
dEQP-VK.wsi.*.incremental_present.*
Components: Vulkan, Framework
VK-GL-CTS issue: 1113
Change-Id: I7ab9fd79c67015e7a1f855c41e0b599481b9d618
Ari Suonpaa [Mon, 9 Apr 2018 05:30:30 +0000 (08:30 +0300)]
Use common utility function for render pass creation
Several tests are creating a simple render pass with only a maximum
of one color and depth/stencil buffer. This task is now done using
an utility function to promote code reusability.
Affects:
dEQP-VK.api.*
dEQP-VK.binding_model.shader_access.*
dEQP-VK.device_group.*
dEQP-VK.draw.shader_viewport_index.*
dEQP-VK.fragment_operations.*
dEQP-VK.geometry.*
dEQP-VK.memory.pipeline_barrier.*
dEQP-VK.pipeline.*
dEQP-VK.protected_memory.*
dEQP-VK.renderpass.sparserendertarget.*
dEQP-VK.robustness.*
dEQP-VK.sparse_resources.*
dEQP-VK.spirv_assembly.instruction.graphics.cross_stage.*
dEQP-VK.synchronization.*
dEQP-VK.tessellation.*
dEQP-VK.ubo.*
dEQP-VK.wsi.*
Components: Vulkan, Framework
VK-GL-CTS issue: 1109
Change-Id: Idc748adf44e0b015b6d0c234d20424f9aec2e9e0
Alexander Galazin [Thu, 19 Apr 2018 10:28:23 +0000 (06:28 -0400)]
Merge "Merge vk-gl-cts/aosp-deqp-dev into vk-gl-cts/master"
Caio Marcelo de Oliveira Filho [Mon, 12 Mar 2018 21:56:57 +0000 (14:56 -0700)]
Check early for PythonInterp 2 in CMakeLists.txt
SPIRV-Tools looks for a Python without specifying the version, which
causes PYTHON_EXECUTABLE to reference a Python 3 executable even if
later code needs PythonInterp 2.
Since VK-GL-CTS currently depends on Python 2, this patch performs
that check early, which will be reused by SPIRV-Tools (i.e. it will
accept python2 found in the system instead of possibly taking a
python3).
Component: Framework
Change-Id: I588c1a1cbe06b95e7fda5db6a1a9f9d046dbda90
Alexander Galazin [Thu, 19 Apr 2018 08:50:40 +0000 (10:50 +0200)]
Merge vk-gl-cts/aosp-deqp-dev into vk-gl-cts/master
Change-Id: I5b8414ca16ddfc52c86d449e01ad23dad32246c6
Dzmitry Malyshau [Fri, 2 Mar 2018 19:22:25 +0000 (14:22 -0500)]
Add CONTRIBUTING.md for Github
Developers used to Github workflow may be tempted to make a PR change,
and it's important to point them to the guidelines as early as possible
in order to avoid the bounced response.
The message they'll see will look like this:
https://help.github.com/articles/setting-guidelines-for-repository-contributors/
Affects: users making Pull Requests on Github.
Components: none
Change-Id: I5921bb4ae4cb8a28f889f3f20c134ca82784c3f2
BMiles [Mon, 19 Mar 2018 18:46:26 +0000 (14:46 -0400)]
This change fixes an issue in BindBufferOffsetEXT
Change hard coded 0 to i for iterating in a loop
The test needs to be iterating through the indices of the buffer.
Affects: KHR-GL4*.transform_feedback.capture_vertex_separate_test
Components: OpenGL
VK-GL-CTS issue: 1108
Change-Id: I0654bf453c05fddad2e10c9843b9130619f4844a
Paavo Pessi [Fri, 6 Apr 2018 11:12:05 +0000 (14:12 +0300)]
Fix image validation errors in SPIR-V tests
A helper function for uploading images caused validation errors by
setting the image layout to SHADER_READ_ONLY_OPTIMAL regardless of the
image descriptor type.
A new parameter for defining the image layout was added to the helper
function, and tests using storage images were changed to use GENERAL
layout.
Affects:
dEQP-VK.spirv_assembly.instruction.compute.image_sampler.imageread.storage_image.*
Components: Vulkan, Framework
VK-GL-CTS issue: 1111
Change-Id: I783c87fe2dd91d65537343d1ae88e3a4f07b77c6
Ari Suonpaa [Mon, 26 Mar 2018 13:22:05 +0000 (16:22 +0300)]
Use common utility function for graphics pipeline creation
Graphics pipeline creation requires several structures to be input. By
replacing the inline implementations of graphics pipeline creation with
a common utility function we can eliminate most of the structures used.
The utility function will use default structure values in these cases.
Affects:
dEQP-VK.fragment_operations.*
dEQP-VK.geometry.*
dEQP-VK.image.*
dEQP-VK.pipeline.*
dEQP-VK.protected_memory.*
dEQP-VK.api.*
dEQP-VK.memory.pipeline_barrier.*
dEQP-VK.binding_model.shader_access.*
dEQP-VK.device_group.*
dEQP-VK.renderpass.*
dEQP-VK.robustness.*
dEQP-VK.glsl.*
dEQP-VK.texture.*
dEQP-VK.rasterization.*
dEQP-VK.ubo.*
dEQP-VK.clipping.*
dEQP-VK.ycbcr.*
dEQP-VK.wsi.*
dEQP-VK.sparse_resources.*
dEQP-VK.spirv_assembly.*
dEQP-VK.subgroups.*
dEQP-VK.synchronization.*
dEQP-VK.tessellation.*
Components: Vulkan, Framework
VK-GL-CTS issue: 1069
Change-Id: I44baccaa8eb670fccf2755e4c3a3370966c3765d
Chris Forbes [Tue, 17 Apr 2018 20:51:08 +0000 (13:51 -0700)]
Merge remote-tracking branch 'khronos/master' into deqp-dev
Change-Id: Ie0f7b6e81907b5b35f9eadf696063d0ec147c168
Alexander Galazin [Tue, 17 Apr 2018 09:44:33 +0000 (11:44 +0200)]
Retarget cts-runner to the master mustpass
Components: OpenGL
Affects: cts-runner
Change-Id: If9a68026a659006dd16d30a1d5ece7d8eb6e4153
Graeme Leese [Mon, 16 Apr 2018 15:04:07 +0000 (16:04 +0100)]
Use layered image bindings correctly
Where the shader uses, for example, image3d the image binding in the API
must be layered, otherwise behaviour is undefined.
Components: OpenGL
Affects: KHR-GLES31.shaders.layout_location.*
VK-GL-CTS Issue: 1123
Change-Id: I8267428d847d426945f1356bd654fb4f6c374881
Alexander Galazin [Tue, 17 Apr 2018 09:10:55 +0000 (11:10 +0200)]
Merge vk-gl-cts/opengl-es-cts-3.2.5 into vk-gl-cts/master
Change-Id: I8ed9b99ed838ef203a3af2c3bbfb769733a7e04e
Chris Forbes [Mon, 16 Apr 2018 21:51:58 +0000 (14:51 -0700)]
Merge remote-tracking branch 'aosp/master' into deqp-dev
Change-Id: I04266dedcbdbdea1b4e6d87bf662ba3d9317f411
Jeff Leger [Thu, 1 Feb 2018 01:31:13 +0000 (20:31 -0500)]
Relax the precision requirements for derivate built-ins
Allow for relaxed precision thredholds for derivatives,
with a quality warning.
Affects: dEQP-GLES3.functional.shaders.derivate.*
Components: AOSP
Khronos vk-gl-cts issue: 977
Change-Id: I1d69ad71b79a30a36daf2bf75b2f50d309bc5496
Jesse Hall [Mon, 4 Dec 2017 21:49:31 +0000 (13:49 -0800)]
Use NDK's CMake toolchain for NDK r15+
The ndk-r11.cmake file is pretty complex because neither CMake nor the
NDK provided adequate support. Starting in NDK r15, the NDK provides
an officially supported CMake toolchain file. So dEQP's
ndk-modern.cmake, used for NDK r15 and later, can just delegate to that
for everything that isn't specific to dEQP, instead of duplicating
large parts of it. This will make future NDK versions easier to support.
Components: Framework, AOSP
VK-GL-CTS issue: 723
Google bug:
75980403
Change-Id: Icc6ec38601017d2fb6ca40defe2f8e73ba76ca96