Jeff Leger [Sun, 22 Oct 2017 15:54:31 +0000 (11:54 -0400)]
Relax precision for atan2 built-in precision tests
The precision of atan2 is relaxed to 5 ULPs.
Affects:
dEQP-VK.glsl.builtin.precision.atan2.*
Components: Vulkan
VK_GL_CTS issue: 749
Change-Id: I146226019f77910d1401903f6e89292ff94c089d
Mark Adams [Mon, 4 Dec 2017 22:26:54 +0000 (17:26 -0500)]
Don't assume rendering complete before buffer latch
It's valid for the buffer to latch before rendering is completed, and
this is causing intermittent failures.
Bug: b/
69981429
Test: --deqp-case=dEQP-EGL*get_frame_timestamps*
(cherry picked from commit
9e0214cf61df1109927f8f23dd5749735057eb30)
Change-Id: Ia9f1d0e4ffa7f3e16a46fc4566157a651d1b0601
Piotr Byszewski [Fri, 3 Nov 2017 09:11:09 +0000 (10:11 +0100)]
Block/non-block uniforms match
Test if uniforms in anonymous blocks do not match loose uniforms with
the same name in a different stage.
Components: OpenGL
VK-GL-CTS issue: 707
Affects:
KHR-GLES3.shaders.uniform_block.common.name_matching
KHR-GL33.shaders.uniform_block.common.name_matching
Change-Id: Ib1ea850ffbea7eb62ddbd6008463c8855f764c7e
Jari Komppa [Tue, 21 Nov 2017 12:52:23 +0000 (14:52 +0200)]
Added tests for different shader varying vec width
Vulkan specification (1.0.8, 14.1.3) states that vertex and fragment
shader varyings may differ as long as they are of the same type and the
vertex shader outputs at least as wide vectors as the fragment shader
inputs.
This set of tests link various varying vector types, all of which
are legal even though they do not match.
Affects:
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_vec4_vec3
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_vec4_vec2
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_vec4_float
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_vec3_vec2
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_vec3_float
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_vec2_float
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_ivec4_ivec3
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_ivec4_ivec2
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_ivec4_int
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_ivec3_ivec2
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_ivec3_int
dEQP-VK.glsl.linkage.varying.rules.legal_mismatch_ivec2_int
Components: Vulkan
VK-Gl-CTS issue: 821
Change-Id: I4afc3edb77fd6f421cafce20d09bbdd948f16dcf
Paavo Pessi [Tue, 10 Oct 2017 08:24:39 +0000 (11:24 +0300)]
Test binding model with multiple descriptor sets
Tests were added to verify that multiple descriptor sets can be used
and accessed in all possible shader stages with all possible descriptor
types.
Multiple descriptor sets are updated using
- vkUpdateDescriptorSets
- vkUpdateDescriptorSetWithTemplateKHR
DescriptorSetUpdateBuilder utility class was modified slightly to
work better with multiple descriptor sets.
New tests:
dEQP-VK.binding_model.shader_access.*.*.*.multiple_descriptor_sets.*
Components: Vulkan, Framework
VK-GL-CTS issue: 91
Change-Id: I6c1a719a7a2af74c3f1eb2a620af1ec73c113411
Andrey Tuganov [Tue, 28 Nov 2017 23:28:18 +0000 (18:28 -0500)]
Fix OpCopyObject from U32 to S32
Conversion from U32 to S32 can use neither OpCopyObject (different types),
nor OpSConvert (same bit width). Changed to OpBitcast.
Affects:
dEQP-VK.spirv_assembly.instruction.compute.indexing
Components: Vulkan
VK-GL-CTS issue: 874
Change-Id: I298c2687f3f6f08f28f928623a1002ec72a5a8f8
Chad Versace [Fri, 3 Nov 2017 16:36:30 +0000 (09:36 -0700)]
Fix build when .git is a gitlink file
If ${CMAKE_SOURCE_DIR}/.git exists, then
framework/qphelper/CMakeLists.txt uses the git revision as the dEQP
revision. The CMake rules update the revision when the files
${CMAKE_SOURCE_DIR}/.git/HEAD and ${CMAKE_SOURCE_DIR}/.git/index
change.
However, Git does not require that .git be a directory. If
git-submodules or git-worktrees are used, then .git may be a gitlink
file instead and the real git directory is elsewhere.
Fix the build by using `git rev-parse --git-dir` to find the real git dir
and `git rev-parse HEAD` to query the commit sha1.
This bumps the CMake requirement from 2.8 to 2.8.2. But that should be
safe since 2.8.2 was released in 2010.
Components: Framework
Change-Id: I811630e5afada8e6aed611b0ef6c51e39be2200b
Piotr Byszewski [Thu, 12 Oct 2017 08:24:48 +0000 (10:24 +0200)]
Test required_internalformat
GTFExtensionTestRequiredInternalformat was ported to glcts and
highly modified to resolve found issues and enable possibility
to also test core GL.
Components: OpenGL
VK-GL-CTS issue: 718
Affects:
KHR-GL40.internalformat.*
KHR-GLES2.core.internalformat.*
KHR-GLES3.core.internalformat.*
KHR-GLES31.core.internalformat.*
KHR-GLES32.core.internalformat.*
Change-Id: Iaf559e7c3fe955e314402c5190523b2c57c7041e
Courtney Goeltzenleuchter [Tue, 27 Jun 2017 20:09:01 +0000 (14:09 -0600)]
Vulkan: Add wide-color tests
Affects:
dEQP-VK.wsi.*.colorspace.*
Components: Vulkan
Change-Id: I6813afa38e320ce91642802dfd0c5b6f2d5543b9
Alexander Galazin [Thu, 30 Nov 2017 13:39:21 +0000 (14:39 +0100)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: If36eca15c35d7c5b473b0bcd6af6c40ef0e68322
Paavo Pessi [Fri, 24 Nov 2017 09:03:06 +0000 (11:03 +0200)]
Fix stageMask (0x1) usage in binding_model and glsl tests
Pipeline barriers using PIPELINE_STAGE_TOP_OF_PIPE_BIT with nonzero
access mask were modified to use proper pipeline stage flags.
Memory barrier structure members were renamed to match the current
specification.
Affects:
dEQP-VK.binding_model.*
dEQP-VK.glsl.derivate.*.texture.*
dEQP-VK.glsl.discard.*_texture
dEQP-VK.glsl.texture_functions.*
dEQP-VK.glsl.texture_gather.*
dEQP-VK.glsl.builtin_var.fragdepth.*
dEQP-VK.glsl.builtin.*
dEQP-VK.glsl.opaque_type_indexing.*
dEQP-VK.glsl.atomic_operations.*
VK-GL-CTS issue: 806
Change-Id: Ia223039abed5c4b15ae004fbaa9fba1ebe92056e
Mikael Tuomisalo [Fri, 17 Nov 2017 13:54:41 +0000 (15:54 +0200)]
Add new tests for OpSConvert and OpFConvert
Added missing tests for OpSConvert and OpFConvert for compute and
graphics shaders.
Edited SPIR-V templates for OpSpecConstantOp-tests to support data types
other than int32 on tests where other types are required.
New tests:
dEQP-VK.spirv_assembly.instruction.compute.opspecconstantop.sconvert
dEQP-VK.spirv_assembly.instruction.compute.opspecconstantop.fconvert
dEQP-VK.spirv_assembly.instruction.graphics.opspecconstantop.sconvert*
dEQP-VK.spirv_assembly.instruction.graphics.opspecconstantop.fconvert*
Components: Vulkan
VK-GL-CTS issue: 711
Change-Id: Ia3cc3bc3f70f5216e364a2da4fb8fe78079fa445
Jari Komppa [Wed, 8 Nov 2017 11:05:44 +0000 (13:05 +0200)]
Removed duplicate fvec3 declaration
A couple of tests had duplicate declarations that resolved into
a vector of three floats.
Affects:
dEQP-VK.spirv_assembly.instruction.compute.opphi.vartype_vec3
dEQP-VK.spirv_assembly.instruction.compute.opphi.vartype_struct
Components: Vulkan
VK-GL-CTS issue: 808
Change-Id: I2b2c865f138ebf0fe67fd3b72d413c040c335170
Paavo Pessi [Tue, 14 Nov 2017 12:56:41 +0000 (14:56 +0200)]
Fix PIPELINE_STAGE_TOP_OF_PIPE_BIT usage in api tests
Test cases using PIPELINE_STAGE_TOP_OF_PIPE_BIT with nonzero access
mask were modified to use proper pipeline stage flags.
The dstAccessMask mask of a buffer memory barrier created by
ComputeInstanceResultBuffer class was changed from
VK_ACCESS_SHADER_WRITE_BIT to VK_ACCESS_HOST_READ_BIT to better
represent its usage.
Memory barrier structure members were renamed to match the current
specification.
Affects:
dEQP-VK.api.buffer_view.access.*
dEQP-VK.api.command_buffers.render_pass_continue
dEQP-VK.api.command_buffers.order_bind_pipeline
dEQP-VK.api.copy_and_blit.core.image_to_buffer.*
dEQP-VK.api.copy_and_blit.core.buffer_to_buffer.*
dEQP-VK.api.copy_and_blit.dedicated_allocation.image_to_buffer.*
dEQP-VK.api.copy_and_blit.dedicated_allocation.buffer_to_buffer.*
VK-GL-CTS issue: 806
Change-Id: Ifaff60b1cd615889fa4b1fc175a78bdbc00be50c
Piotr Byszewski [Thu, 9 Nov 2017 09:49:25 +0000 (10:49 +0100)]
Test if mediump optimizations are not to aggressive
GTF sin and cos tests were ported to glcts as they can also
be used to detect overly optimized set of mediump operations.
Components: OpenGL
VK-GL-CTS issue: 706
Affects:
KHR-GL43.aggressive_optimizations.*
KHR-GLES2.shaders.aggressive_optimizations.*
KHR-GLES3.shaders.aggressive_optimizations.*
KHR-GLES31.shaders.aggressive_optimizations.*
KHR-GLES32.shaders.aggressive_optimizations.*
Change-Id: Idd2ea75ea552b9106792775570d64b347db32276
Paavo Pessi [Mon, 30 Oct 2017 13:25:49 +0000 (15:25 +0200)]
Test component qualifier in shaders
Tests were added to verify that various combinations of component
qualifiers can be used with different data types (vec2, vec3, vec4,
ivec2 and ivec4) in following scenarios:
- in vertex shader input
- in fragment shader output
- when passing data from vertex shader to fragment shader
New tests:
dEQP-VK.glsl.440.linkage.varying.component.*
Components: Vulkan
VK-GL-CTS issue: 741
Change-Id: Ie18c420b9d3ba8bd082b98f4618d7fc595b8baba
Alexander Galazin [Thu, 23 Nov 2017 11:27:18 +0000 (12:27 +0100)]
Delete GTF mustpass files
OpenGL ES CTS no longer uses GTF module,
time to delete the source files for mustpass
Components: OpenGL
Change-Id: Ib234e47612e48b50e6c34544cebbaba11a2f08c3
Piotr Byszewski [Tue, 31 Oct 2017 16:57:52 +0000 (17:57 +0100)]
Correct robust buffer access behavior tests
Robust buffer access behavior tests vere moved to contexless module.
In addition common version of those test was generalized so that it
can test both es and core GL.
Components: OpenGL
VK-GL-CTS issue: 793
Affects:
KHR-NoContext.es32.robust_buffer_access_behavior.*
KHR-NoContext.gl43.robust_buffer_access_behavior.*
Change-Id: Idd967d00635faff4ec41d57abf203f22286bc145
Jari Komppa [Tue, 14 Nov 2017 12:21:06 +0000 (14:21 +0200)]
Added constraints for random access flags
The pipeline barrier tests randomize access flags and stages, sometimes
resulting in access flags being enabled for stages that do not support
said accesses. This fix removes the illegal access flags while leaving
the random flag generation intact.
Affects:
dEQP-VK.memory.pipeline_barrier.*
Components: Vulkan
VK-GL-CTS issue: 814
Change-Id: I135ea8a9e41b5d8428ef77fe6ecbdf2380d80c7d
Juan A. Suarez Romero [Fri, 10 Nov 2017 11:20:36 +0000 (11:20 +0000)]
Do not use depthClampEnable unconditionally.
Depth clamping is not needed by these tests, so
do not rely on the optional feature when creating
the pipeline object.
Affects:
dEQP-VK.spirv_assembly.*
Components: Vulkan
VK-GL-CTS issue: 846
Change-Id: I248759e6ac209ccbd7661068c3ef5f1dfb8ca501
Juan A. Suarez Romero [Fri, 10 Nov 2017 09:50:10 +0000 (09:50 +0000)]
Remove duplicated definitions in spirv_assembly tests
Duplicated definitions are now allowed.
Affects:
dEQP-VK.spirv_assembly.instruction.compute.opphi.vartype_*
dEQP-VK.spirv_assembly.instruction.compute.opspecconstantop.vector_related
dEQP-VK.spirv_assembly.instruction.graphics.opspecconstantop.vector_related_*
Components: Vulkan
VK-GL-CTS issue: 847
Change-Id: Ie09293ece03eeb25893aa66e7cb23e7ad8fa5e95
Slawomir Cygan [Fri, 27 Oct 2017 10:54:54 +0000 (12:54 +0200)]
Test with depth range greater than 1.0 should require extension
"_deltalarge" variant of tests was using minDepth/maxDepth
outside of valid usage from unextended spec.
Added requirement for VK_EXT_depth_range_unrestricted.
Affects:
dEQP-VK.draw.inverted_depth_ranges.*
Components: Vulkan
VK-GL-CTS issue: 795
Change-Id: I18f1c73d7250865aaba5ed64e6b8da955fec0aea
Alexander Galazin [Wed, 15 Nov 2017 14:46:11 +0000 (15:46 +0100)]
Merge vk-gl-cts/opengl-cts-4.6.0 into vk-gl-cts/master
Change-Id: I6b148f6e9948675fbe51b5e7ee04ec4cca24df2c
Ari Suonpaa [Fri, 1 Sep 2017 10:40:33 +0000 (13:40 +0300)]
Add testing for sparse D/S/DS images.
Add sparse binding support for shadow texture tests by supporting sparse
binding in texture test util class. Removed unnecessary image aspect
check which prevented shadow sampler testing with sparse images.
Affects:
dEQP-VK.glsl.texture_functions.texture.sparse*shadow*
dEQP-VK.glsl.texture_gather.*depth32f*sparse*
dEQP-VK.texture.shadow.*sparse*
Components: Vulkan
VK-GL-CTS issue: 79
Change-Id: I265f09817493d96b91450f71822b00434bcf6cf3
Alexander Galazin [Wed, 8 Nov 2017 08:53:26 +0000 (09:53 +0100)]
Notice of withdrawal of Vulkan CTS 1.0.0.x and 1.0.1.x
Components: Vulkan
Change-Id: I87ec80e470e72e9a51760fdb1f441099b9c4d3a8
Alexander Galazin [Fri, 10 Nov 2017 15:13:33 +0000 (16:13 +0100)]
Remove mod from GL mustpass
Components: OpenGL
VK-GL-CTS public issue: 51
Affects:
KHR-GL*.gpu_shader_fp64.builtin.mod_*
Change-Id: Idac62351b9ca926cdcb411990c33e7332e8a9db7
Alexander Galazin [Fri, 10 Nov 2017 12:11:07 +0000 (13:11 +0100)]
Update KC CTS reference
Components: OpenGL
Change-Id: I241503a713fc4a15cdcf359c40e2b119b5b5fd4f
Adam Czupryna [Thu, 26 Oct 2017 09:56:11 +0000 (11:56 +0200)]
Change Built-in Functions test case proposal
Remove build-in noise* functions testing as tey are deprecated.
Section GLSL to SPIR-V Features Tests / Built-in Functions
Components: OpenGL
VK-GL-CTS issue: 554
Change-Id: I0f3d19bd6fd70c54a32610e68250f4f39f14a708
Juan A. Suarez Romero [Mon, 30 Oct 2017 12:50:23 +0000 (12:50 +0000)]
Use proper name for array-of-array varyings.
varying_structure_location tests have the piece of shader code:
```
struct Data {
dvec2 single;
dvec2 array[1];
};
layout (location = 0) flat in Data tes_gs_output[][1];
```
When test tries to verify "single" variable, it tries to access through
"tes_gs_output[0].single", which is not correct. Rather it must use
"tes_gs_output[0][0].single"
Affects:
* KHR-GL46.enhanced_layouts.varying_structure_locations
Components: OpenGL
VK-GL-CTS issue: 796
Change-Id: Ib1c2ac6877f46f0be8530156a184b8fa41349c40
Igor Ostrowski [Thu, 19 Oct 2017 16:41:40 +0000 (18:41 +0200)]
Add depth bounds tests
New tests:
dEQP-VK.pipeline.depth.format.*_depth_bounds_test
dEQP-VK.dynamic_state.ds_state.depth_bounds_2
Rename:
dEQP-VK.dynamic_state.ds_state.depth_bounds -> depth_bounds_1
Components: Framework, Vulkan
VK-GL-CTS issue: 724
Change-Id: Ia3efb59c0d295f96687d4e63e34f882286cb41f6
Juan A. Suarez Romero [Mon, 6 Nov 2017 18:50:28 +0000 (19:50 +0100)]
GL_ARB_compute_shader required for OpenGL < 4.3
GL_ARB_compute_shader is part of OpenGL 4.3 and above.
For older versions, it is required to enable it explicitly.
Fixes:
KHR-GL31.CommonBugs.CommonBug_ParenthesisInLayoutQualifierIntegerValue
KHR-GL32.CommonBugs.CommonBug_ParenthesisInLayoutQualifierIntegerValue
KHR-GL33.CommonBugs.CommonBug_ParenthesisInLayoutQualifierIntegerValue
KHR-GL40.CommonBugs.CommonBug_ParenthesisInLayoutQualifierIntegerValue
KHR-GL41.CommonBugs.CommonBug_ParenthesisInLayoutQualifierIntegerValue
KHR-GL42.CommonBugs.CommonBug_ParenthesisInLayoutQualifierIntegerValue
Components: OpenGL
VK-GL-CTS issue: 818
Change-Id: I8e7ab78e29861547f0379feebce0375e1dabf47c
Alexander Galazin [Fri, 10 Nov 2017 10:24:21 +0000 (11:24 +0100)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: Ib7a8d1973b881cb59b750ac9e2c291bc2602f559
Arkadiusz Sarwa [Tue, 10 Oct 2017 15:35:42 +0000 (17:35 +0200)]
Test local workgroup size
Add tests:
dEQP-VK.spirv_assembly.instruction.compute.localsize*
VK-GL-CTS issue: 629
Compoonents: Vulkan
Change-Id: Ibf3a4db7058c99c2dfa498a7c3ddd8794c0af294
Igor Ostrowski [Thu, 26 Oct 2017 13:21:18 +0000 (15:21 +0200)]
Fix issues reported by validation layers
Affects:
dEQP-VK.draw.*
dEQP-VK.query_pool.*
dEQP-VK.dynamic_state.*
Components: Vulkan
VK-GL-CTS issue: 809
Change-Id: I6d8d227612e9cf30ad08921e922d756e0fa5c78d
Jorg Wagner [Mon, 6 Nov 2017 13:56:49 +0000 (14:56 +0100)]
Do not use depthClampEnable unconditionally.
Depth clamping is not needed by this test, so
do not rely on the optional feature when creating
the pipeline object.
Affects:
dEQP-VK.api.smoke.*
Components: Vulkan
VK-GL-CTS issue: 817
Change-Id: I6794f7f5d17d7610ff43dcb4080a6a30e12ed95b
Slawomir Cygan [Tue, 31 Oct 2017 10:42:20 +0000 (11:42 +0100)]
Pipeline used in second subpass should set subpass index to 1
Affects:
dEQP-VK.renderpass.sampleread
Components: Vulkan
VK-GL-CTS issue: 800
Change-Id: If387b43ac8469544f9270d747742cbcec66a1f5c
Juan A. Suarez Romero [Tue, 7 Nov 2017 12:24:25 +0000 (12:24 +0000)]
Load functions from ARB_uniform_buffer_object
While this extension is part of OpenGL 3.1 and above, it is not part of
OpenGL 3.0.
As we need to run the tests for OpenGL 3.0 too, and we need to use some
functions exposed by this extension, let's load them from the extension
itself, instead of relying in the core functions.
Affects:
* KHR-GL30.transform_feedback.draw_xfb_instanced_test
* KHR-GL30.transform_feedback.draw_xfb_stream_instanced_test
Components: OpenGL
VK-GL-CTS issue: 819
Change-Id: I2cd1652b4c2d3b55e009c44b85b451d207387b96
Ari Suonpaa [Fri, 29 Sep 2017 07:41:00 +0000 (10:41 +0300)]
Add tests for verifying standard image block shape usage.
Added tests that verify the sparse image block shape
requirements match with standard block shapes in case
the use of standard block shapes is advertised.
Affects:
dEQP-VK.sparse_resources.image_block_shapes.*
Components: Vulkan
VK-GL-CTS issue 101
Change-Id: Icb0d15c15d437ae36a6c5e92e8d454ded06f662e
Courtney Goeltzenleuchter [Tue, 12 Sep 2017 23:11:21 +0000 (17:11 -0600)]
Catch BAD_MATCH from createWindowSurface
On some platforms, it's possible to fail createWindowSurface
for some attribute combinations. Catch that case and return
NotSupported rather than a test failure.
Test: adb -d shell am start \
-n com.drawelements.deqp/android.app.NativeActivity \
-e cmdLine '"deqp --deqp-case=dEQP-EGL.functional.wide_color.* \
--deqp-log-filename=/sdcard/dEQP-Log.qpa"'
Bug:
65376495
(cherry picked from commit
7df6c12492c6baf9fdf9a3ea0b7f51e33722a4e8)
Change-Id: I6546cbdfc011c7d26cfa3cdad7cc6f0df1b67122
Mark Adams [Fri, 20 Oct 2017 13:49:08 +0000 (09:49 -0400)]
Multiple draw buffers in blend_equation_advanced
GL_NVX_blend_equation_advanced_multi_draw_buffers adds support for
multiple draw buffers with advanced blend equations. This change
modifies the below test to check for this extension and to expect
GL_NO_ERROR if it is present.
Components: AOSP
VK-GL-CTS issue: 770
Affects:
dEQP-GLES31.functional.debug.negative_coverage.get_error.advanced_blend.attachment_advanced_equation
(cherry picked from commit
2cb35a235c1a7bd3c63b296c37e186eec34e08e2)
Change-Id: I70ba495d96b4a17bed589e3affa693f572fde446
James Helferty [Thu, 19 Oct 2017 18:24:35 +0000 (14:24 -0400)]
Relax XFB begin check to match recent spec edit
Update tests to reflect a recent change to the GL 4.6 specification.
Previously, if a shader layout qualifier specified that xfb_stride was
non-zero for a given xfb_buffer, then a buffer object was required to
be bound at that transform feedback buffer binding point. After the
spec change, it is now legal to not have a buffer bound, provided no
outputs are associated with the buffer.
Components: OpenGL
VK-GL-CTS issue: 785
Affects:
* KHR-GL44/45/46.enhanced_layouts.xfb_stride_of_empty_list
* KHR-GL44/45/46.enhanced_layouts.xfb_stride_of_empty_list_and_api
Change-Id: I38622bad1af3bc8a2a4c683eadc184e6231f7c85
Adam Czupryna [Fri, 7 Jul 2017 09:42:37 +0000 (11:42 +0200)]
Add CTS_ARB_gl_spirv test implementation
Add CTS_ARB_gl_spirv test implementation that contains:
SpirvModulesPositiveTest,
SpirvModulesShaderBinaryMultipleShaderObjectsTest,
SpirvModulesStateQueriesTest,
SpirvModulesErrorVerificationTest,
SpirvGlslToSpirVEnableTest,
SpirvGlslToSpirVBuiltInFunctionsTest,
SpirvGlslToSpirVSpecializationConstantsTest,
SpirvValidationBuiltInVariableDecorationsTest
Affects:
KHR-GL45.gl_spirv.*
Components: Framework, OpenGL
VK-GL-CTS issue: 554
Change-Id: I0d54c70df619dbbd80b28350d2ac5e31aca08927
(cherry picked from commit
a24448cdd72ffdbd8f7f571886625b8a53100979)
Adam Czupryna [Fri, 7 Jul 2017 09:42:37 +0000 (11:42 +0200)]
Add CTS_ARB_gl_spirv test implementation
Add CTS_ARB_gl_spirv test implementation that contains:
SpirvModulesPositiveTest,
SpirvModulesShaderBinaryMultipleShaderObjectsTest,
SpirvModulesStateQueriesTest,
SpirvModulesErrorVerificationTest,
SpirvGlslToSpirVEnableTest,
SpirvGlslToSpirVBuiltInFunctionsTest,
SpirvGlslToSpirVSpecializationConstantsTest,
SpirvValidationBuiltInVariableDecorationsTest
Affects:
KHR-GL45.gl_spirv.*
Components: Framework, OpenGL
VK-GL-CTS issue: 554
Change-Id: I0d54c70df619dbbd80b28350d2ac5e31aca08927
Chris Forbes [Mon, 23 Oct 2017 22:13:38 +0000 (15:13 -0700)]
KHR_swapchain: Trigger the fence off vkQueueSubmit
Another instance of
438ea0888e8ccae.
VK-GL-CTS issue: 756
Component: Vulkan
Affects:
dEQP-VK.wsi.*.swapchain.*
Change-Id: I87da6800e0c8402c2c50ca88eb3360b31c3a515d
Chris Forbes [Mon, 23 Oct 2017 22:09:56 +0000 (15:09 -0700)]
GOOGLE_display_timing: Trigger the fence off vkQueueSubmit
Another instance of
438ea0888e8ccae.
VK-GL-CTS issue: 756
Component: Vulkan
Affects:
dEQP-VK.wsi.*.display_timing.*
Change-Id: If94644402579ba1f46347d155313f6ef52d125a0
Ari Suonpaa [Wed, 4 Oct 2017 12:10:29 +0000 (15:10 +0300)]
Add tests for aligned mip size in sparse images.
Added tests that verify the mip tail starts at the correct LOD
in case aligned mip size is enabled.
Affects:
dEQP-VK.sparse_resources.aligned_mip_size.*
Components: Vulkan
VK-GL-CTS issue 101
Change-Id: I07f1723359dc20ad2edb18ac668ea5b0fd6785be
Alexander Galazin [Tue, 31 Oct 2017 14:19:33 +0000 (15:19 +0100)]
Merge vk-gl-cts/opengl-cts-4.6.0 into vk-gl-cts/master
Change-Id: I92cf294d57f1ea0f22d8d65a0a246984612952ae
Alexander Galazin [Tue, 31 Oct 2017 14:19:33 +0000 (15:19 +0100)]
Merge vk-gl-cts/opengl-es-cts-3.2.4 into vk-gl-cts/master
Change-Id: I6875bda7e1c94df9873486233180d1ce158f302a
Alexander Galazin [Tue, 31 Oct 2017 11:42:31 +0000 (12:42 +0100)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: I74b2938bb1e09131e2e3e2c22515e18ed3970138
Jason Ekstrand [Thu, 19 Oct 2017 16:07:52 +0000 (09:07 -0700)]
VK_KHR_variable_pointers: Don't use OpPtrAccessChain on a bare SSBO
You can't use OpPtrAccessChain on the direct result of an OpVariable
because it's not the element of any array. Instead, we need to do our
OpPtrAccessChain on a pointer to an actual array element and that
pointer needs to be properly decorated with a stride.
Component: Vulkan
VK-GL-CTS issue: https://gitlab.khronos.org/Tracker/vk-gl-cts/issues/686
Affects:
dEQP-VK.spirv_assembly.instruction.graphics.variable_pointers.multi_buffer_*
Change-Id: I9c45fb1320099e404672ee169a6a8e5c41eb13f7
Jason Ekstrand [Thu, 19 Oct 2017 15:43:36 +0000 (08:43 -0700)]
VK_KHR_variable_pointers: Delete an unused and invalid variable
The %float_var variable was not being used in any way by the test. It
was also invalid because it had the StorageBuffer storage class (so it's
in external memory) but also had a constant initializer. The real
variable being tested in these tests is %f32_ptr_var which has the
Function storage class so a constant initializer is fine.
Component: Vulkan
VK-GL-CTS issue: https://gitlab.khronos.org/Tracker/vk-gl-cts/issues/686
Affects:
dEQP-VK.spirv_assembly.instruction.graphics.variable_pointers.nullptr_graphics.opvariable_initialized_null_*
dEQP-VK.spirv_assembly.instruction.graphics.variable_pointers.nullptr_graphics.opselect_null_or_valid_ptr_*
Change-Id: I9e2f118b1c774a84c0b6900ef0db081cb5256c2f
Iago Toral Quiroga [Mon, 2 Oct 2017 11:10:20 +0000 (13:10 +0200)]
Fix attribute mapping in enhanced layouts tests for input arrays
Some tests in KHR-GL45.enhanced_layouts.varying_* (specifically
KHR-GL45.enhanced_layouts.varying_array_{locations,components})
use input arrays on the vertex shader, however, the code
that uploads vertex data for these (and other test cases) doesn't
account for arrays at all.
Also, the tests in this category that call
TextureTestBase::prepareAttributes to do the VAO setup seem to
provide inconsistent values for variable.m_descriptor.m_offset,
which also leads to incorrect vertex attribute mappings.
Specifically, in the 'component' variants of these tests
(such as varying_array_components), two different input variables
have m_offset = 0, which has the consequence that we end up
computing a vertex buffer with a size that's only big enough to
fit one of the inputs and we overwrite the data for the first
input with the data for the second input when we setup the VAO,
however, the 'location' variants (such as varying_array_locations),
see different offsets for different input variables, which leads to
a different VAO setup.
Components: OpenGL
VK-GL-CTS issue: 739
Affects:
KHR-GL45.enhanced_layouts.varying_*
KHR-GL46.enhanced_layouts.varying_*
Change-Id: Icec238a970e2cd67d6109a25fba980f756abb71b
Juan A. Suarez Romero [Wed, 18 Oct 2017 09:37:27 +0000 (09:37 +0000)]
Several fixes for texture_size_promotion
Specifically, this changeset fixes:
- Check for unsupported framebuffer configurations
- Setup data for depth/stencil textures
- Explicitly set minification/magnification filters
- Deal with the corner cases in precision when using fixed-point
textures.
Affects:
KHR-GL45.texture_size_promotion.functional
KHR-GL46.texture_size_promotion.functional
Components: OpenGL
VK-GL-CTS issue: 783
Change-Id: I3f191084b2a604f1a743759905d641d2b13db2ad
Iago Toral Quiroga [Tue, 17 Oct 2017 07:58:57 +0000 (09:58 +0200)]
Respect GL_MAX_TRANSFORM_FEEDBACK_INTERLEAVED_COMPONENTS
KHR-GL45.enhanced_layouts.xfb_override_qualifiers_with_api
generates shaders that capture XFB for varyings of all types.
For the cases of dmat4x3, dmat3x4 and dmat4, this requires
over 64 transform feedback components in total, which is larger
than the minimum value that implementations are required to
support for GL_MAX_TRANSFORM_FEEDBACK_INTERLEAVED_COMPONENTS,
which the OpenGL 4.6 spec sets in 64. For implementations that
support this minimum, the generated shaders are invalid and
expected to fail to link.
Check the required component size of the XFB for the generated
shaders and skip them if they exceed the implementation limit.
At worse, for implementations that only support the minimum
number of transform feedback components required by the spec
it would only skip the 3 largest types (out of 34 being tested).
Components: OpenGL
VK-GL-CTS issue: 776
Affects:
KHR-GL45.enhanced_layouts.xfb_override_qualifiers_with_api
Change-Id: I51c9da494f96a0746f74116f817d99afd070a9ee
Iago Toral Quiroga [Wed, 18 Oct 2017 10:08:54 +0000 (12:08 +0200)]
Component aliasing on vertex inputs is allowed
From the OpenGL 4.6 spec, section 11.1 Vertex shaders,
page 384:
"It is possible for an application to bind more than one
attribute name to the same location. This is referred
to as aliasing. This will only work if only one of the
aliased attributes is active in the executable program,
or if no path through the shader consumes more than one
attribute of a set of attributes aliased to the same
location. A link error can occur if the linker determines
that every path through the shader consumes multiple
aliased attributes, but implementations are not required
to generate an error in this case."
The varying_location_aliasing_with_mixed_types enhanced
layouts test generates negative tests to check that
component aliasing produces linker errors for all shader
stages, including vertex shaders when the aliasing occurs
on the vertex inputs. By the citation above, the spec
doesn't mandate implementations to produce an error in
this case, so these particular tests are invalid.
This change removes the test cases that expect an error
to be generated for shaders with component aliasing on
vertex shader inputs.
Components: OpenGL
VK-GL-CTS issue: 781
Affects:
KHR-GL45.enhanced_layouts.varying_location_aliasing_with_mixed_types
KHR-GL46.enhanced_layouts.varying_location_aliasing_with_mixed_types
Change-Id: Ie1cafc92c4b7023c3d911ec5fe883dad297029d5
Eleni Maria Stea [Tue, 13 Jun 2017 07:08:08 +0000 (09:08 +0200)]
Online compression for ETC2 compressed formats is not required
The DSA compressed texture tests currently require online compression for
ETC2 formats but drivers are not required to support this. Since the purpose
of the test is not to test the compression, when a compression format is not
supported we use already compressed data.
Components: OpenGL
VK-GL-CTS issue: 497
Affects:
KHR-GL45.direct_state_access.textures_compressed_subimage
Change-Id: Icff52745e9cf7a2acb39aea71d31f4ff827ea01c
Kenneth Graunke [Thu, 12 Oct 2017 04:39:05 +0000 (21:39 -0700)]
Delete incorrect invalid map transform feedback subtest.
The citation given in this test comes from the OpenGL man pages, which
are not authoritative. The ES 3.2 and GL 4.x specifications say:
"Effects of Mapping Buffers on Other GL Commands
Any GL command which attempts to read from, write to, or change the
state of a buffer object may generate an INVALID_OPERATION error if
all or part of the buffer object is mapped. However, only commands
which explicitly describe this error are required to do so. If an
error is not generated, using such commands to perform invalid
reads, writes, or state changes will have undefined results and may
result in GL interruption or termination."
Importantly, it says "_may_ generate an INVALID_OPERATION error". Some
drivers choose not to implement this to avoid the CPU overhead required
to enforce this optional restriction.
Affects:
- KHR-*.transform_feedback.api_errors_test
VK-GL-CTS issue: 765
Change-Id: I34ca78926fe0f311db096de98b71e0981ee91d53
Components: OpenGL
Paavo Pessi [Fri, 15 Sep 2017 07:05:52 +0000 (10:05 +0300)]
Test separate and combined images and samplers in SPIR-V
Tests were added to verify that the following Image Instructions
work with related descriptor types in all possible shader stages:
OpImageRead
- Storage image
OpImageFetch
- Sampled image with separate sampler
- Combined image sampler
OpImageSampleExplicitLod
- Sampled image with separate sampler
- Combined image sampler
The following scenarios are covered to verify that images and samplers
work as function parameters and that you can combine an image and a
sampler both before and after passing it into a function.
- Use image and sampler as local variables
- Pass image to function
- Pass sampler to function
- Pass image and sampler to function
New tests:
dEQP-VK.spirv_assembly.instruction.graphics.image_sampler.*
dEQP-VK.spirv_assembly.instruction.compute.image_sampler.*
Components: Vulkan
VK-GL-CTS issue: 104
Change-Id: I1de35fc0099881b3242f6e953a5c9624004e0f3e
Piers Daniell [Tue, 3 Oct 2017 22:08:51 +0000 (16:08 -0600)]
Copy the gles3/graphicsfuzz data to the output dir
Without this there is no gles3/graphicsfuzz data and
the tests fail.
Affects:
dEQP-GLES3.functional.shaders.metamorphic.synthetic.*
VK-GL-CTS issue: 792
Components: OpenGL
Change-Id: Ibb9627afd83d35ab4ae1b02235eb4a65f4cd7722
Jeannot Breton [Fri, 6 Oct 2017 13:56:17 +0000 (08:56 -0500)]
Skip non accelerated pixel format
They were previously skipped in older CTS version,
but when we reorganized the code to merge all CTS together the
related code got remove
We also fix the way we detect hardware acceleration
Affects:
pixel formats selected by cts-runner on Windows
Components: OpenGL
VK-GL-CTS issue: 692
Change-Id: I9968d85b555fe41aea8e64b885f182407e1ce95e
Piotr Byszewski [Fri, 8 Sep 2017 12:51:58 +0000 (14:51 +0200)]
CopyImageSubData with integer textures
Make sure that glCopyImageSubData does not generate
GL_INVALID_OPERATION during copying from a single level integer
texture, with BASE_LEVEL and MAX_LEVEL set to 0 and with the
mipmap filters at the default of GL_NEAREST_MIPMAP_LINEAR and
GL_LINEAR.
Components: OpenGL
VK-GL-CTS issue: 607
Affects:
KHR-GL43.copy_image.integer_tex
KHR-GLES32.core.copy_image.*
Change-Id: Ia160481547fed4266cb51ad0fbec89db4407bf8b
Alexander Galazin [Fri, 20 Oct 2017 10:49:07 +0000 (12:49 +0200)]
Merge vk-gl-cts/opengl-cts-4.6.0 into vk-gl-cts/master
Change-Id: I5a9a2484a8f7e7a70466db2ebd6af4f402279021
Jari Komppa [Tue, 12 Sep 2017 12:00:59 +0000 (15:00 +0300)]
Add tests to use OpPhi with various variable types
Based on the spec, OpPhi should be able to handle any variable type.
These tests use various variable types (int, float, vec3, mat4, array
and struct) with the OpPhi instruction.
All of the tests are of the form:
testtype r
if (indata > 0)
r = 1
else
r = -1
return (float)r
Affects:
dEQP-VK.spirv_assembly.instruction.compute.opphi.vartype_int
dEQP-VK.spirv_assembly.instruction.compute.opphi.vartype_float
dEQP-VK.spirv_assembly.instruction.compute.opphi.vartype_vec3
dEQP-VK.spirv_assembly.instruction.compute.opphi.vartype_mat4
dEQP-VK.spirv_assembly.instruction.compute.opphi.vartype_array
dEQP-VK.spirv_assembly.instruction.compute.opphi.vartype_struct
Components: Vulkan
VK-GL-CTS issue: 259
Change-Id: I2d090d06c52c904d66b35f8cee5f2895a2087069
Chris Forbes [Mon, 16 Oct 2017 19:42:35 +0000 (12:42 -0700)]
Avoid watchdog timeout in Vulkan builtin precision tests
Calculating the reference intervals can take a substantial amount of
time. On some very slow targets, this can cause us to trip the watchdog
in the verification loop.
Same approach and limit as in modules/glshared/glsBuiltinPrecisionTests.cpp
Affected tests:
dEQP-VK.glsl.builtin.precision.*
Component: Vulkan
Bug:
67854695
Test: run on google devices
Change-Id: I492b656e7e2a9aa724bb3785bdf1416a4c22ec04
Ari Suonpaa [Mon, 18 Sep 2017 13:09:52 +0000 (16:09 +0300)]
Add test for reading samples from previous subpass.
Added new test under renderpass group where two subpasses are run
and the second one reads samples from the first. This verifies any
of the samples from an input attachment can be read.
Affects: dEQP-VK.renderpass.sampleread.*
Components: Vulkan
VK-GL-CTS issue: 636
Change-Id: I8863a31cbfade75581e0874c936f87e8b23a1619
Kenneth Graunke [Mon, 25 Sep 2017 04:18:11 +0000 (21:18 -0700)]
Make 420pack 'binding' layout qualifier tests check MAX_*_IMAGE_UNIFORMS
Image uniform support is not mandatory for the VS/TCS/TES/GS stages,
so GL_MAX_*_IMAGE_UNIFORMS can report 0. On such an implementation,
these tests would fail.
This patch skips image testing in any such stages, making them
essentially passthrough shaders.
In BindingImageSingleTest, we also put the binding qualifier in the
fragment shader stage if the stage where we'd normally put it doesn't
support images. This way, it's guaranteed to exist in at least one
shader stage.
Components: OpenGL
VK-GL-CTS issue: 712
Affects:
* KHR-GL45.shading_language_420pack.binding_images
* KHR-GL45.shading_language_420pack.binding_image_api_override
* KHR-GL45.shading_language_420pack.binding_image_array
* KHR-GL45.shading_language_420pack.binding_image_single
* KHR-GL45.shading_language_420pack.binding_image_default
Change-Id: Ic9eceb4472edff87d5de54ec6318cdb856b0bb8d
Kenneth Graunke [Sun, 24 Sep 2017 23:18:11 +0000 (16:18 -0700)]
Skip impossible shader stages in image load store max uniforms test.
Implementations not required to support images in the geometry stages,
and are allowed to advertise 0 for:
* GL_MAX_VERTEX_IMAGE_UNIFORMS
* GL_MAX_TESS_CONTROL_IMAGE_UNIFORMS
* GL_MAX_TESS_EVALUATION_IMAGE_UNIFORMS
* GL_MAX_GEOMETRY_IMAGE_UNIFORMS
This means that gl_Max*ImageUniforms may be 0, which means that
#define N_UNIFORMS gl_MaxVertexImageUniforms
layout(r32i) uniform iimage2D u_image[N_UNIFORMS];
will define a zero sized array, which is illegal. To work around this,
simply skip stages that have zero supported image uniforms - there's
nothing to test there, anyway.
Components: OpenGL
VK-GL-CTS issue: 712
Affects:
* KHR-GL45.shader_image_load_store.uniform-limits
Change-Id: I14116ff8b4714e04b35ce0929ed68ed0f38e7d52
Graeme Leese [Mon, 9 Oct 2017 10:17:28 +0000 (11:17 +0100)]
Broadcom Waiver Request for #757
VK-GL-CTS issue: 757
Affected tests:
dEQP-VK.draw.inverted_depth_ranges.nodepthclamp_deltazero
Component: Vulkan
Change-Id: I22500562b83716697be98605a2b44de83e3132a0
Ari Suonpaa [Thu, 14 Sep 2017 08:02:24 +0000 (11:02 +0300)]
Add tests for inverted depth ranges.
Added a new test under draw group that draws a triangle with
interpolated depth values. Viewport minDepth and maxDepth is
configured so that minDepth >= maxDepth. Test is run with both
depth clamp enabled and disabled.
Affects: dEQP-VK.draw.inverted_depth_ranges.*
Components: Vulkan
VK-GL-CTS issue: 616
Change-Id: I68ff5ea420831600a7cfcc0741820802bb745bdc
Piotr Byszewski [Thu, 14 Sep 2017 09:34:18 +0000 (11:34 +0200)]
Test if GL_FRAGMENT_PRECISION_HIGH is exposed
GL_FRAGMENT_PRECISION_HIGH macro should be exposed in all ES shaders
from version 300 es onwards, as well as in version 100.
Components: OpenGL
VK-GL-CTS issue: 11
Affects:
KHR-GLES3.core.shader_macros.*
KHR-GLES31.core.shader_macros.*
KHR-GLES32.core.shader_macros.*
Change-Id: I8612a5ebe2c565405ae27319d821a0d6cae4fa52
Adam Czupryna [Wed, 4 Oct 2017 12:44:03 +0000 (14:44 +0200)]
Add Shader Binary Multiple Shader Objects test case proposal
This is Shader Binary Multiple Shader Objects test case proposal to
CTS_ARB_gl_spirv specification.
Components: OpenGL
VK-GL-CTS issue: 554
Change-Id: I05817ca5e8daf5f1e08a849a67de3784376f317a
Alexander Galazin [Fri, 13 Oct 2017 07:24:43 +0000 (09:24 +0200)]
Merge vk-gl-cts/opengl-cts-4.6.0 into vk-gl-cts/master
Change-Id: Ief4d74bb43002c377e9bcfa941dd80bc627b68b1
Kenneth Graunke [Fri, 8 Sep 2017 23:18:49 +0000 (16:18 -0700)]
Delete incorrect invalid map CTS tests.
The citation given in these tests comes from the OpenGL man pages, which
are not authoritative. The ES 3.2 and GL 4.x specifications say:
"Effects of Mapping Buffers on Other GL Commands
Any GL command which attempts to read from, write to, or change the
state of a buffer object may generate an INVALID_OPERATION error if
all or part of the buffer object is mapped. However, only commands
which explicitly describe this error are required to do so. If an
error is not generated, using such commands to perform invalid
reads, writes, or state changes will have undefined results and may
result in GL interruption or termination."
Importantly, it says "_may_ generate an INVALID_OPERATION error". Some
drivers choose not to implement this to avoid the CPU overhead required
to enforce this optional restriction.
Affects:
- KHR-*.draw_elements_base_vertex_tests.invalid_mapped_bos
VK-GL-CTS issue: 682
Components: OpenGL
Change-Id: I4c0f7785ca7a747145591d071dda6e5231c787d4
(cherry picked from commit
6a50dd2a2cd0a3f867fe6bdabbf8dc9fb678fb99)
Alexander Galazin [Wed, 11 Oct 2017 10:43:53 +0000 (12:43 +0200)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: I630355a8c937c0fcd2ea102e5b1dcce81eb58753
Piotr Byszewski [Fri, 25 Aug 2017 11:39:58 +0000 (13:39 +0200)]
Add multiple contexts tests
multiple_contexts test (one of shader_subroutine tests) was modified
and moved to contextless module. Shared contexts functionality was
restored in framework.
Components: Framework, OpenGL
VK-GL-CTS issue: 612
Affects:
KHR-NoContext.gl40.multiple_contexts.uniform_preservation
KHR-GL40.shader_subroutine.multiple_contexts
Change-Id: Id170ef37c2e3c488095663c91788978223d356ec
(cherry picked from commit
1b91247d4ad206cbdbef731a2bac9151d6ddda70)
Piotr Byszewski [Thu, 7 Sep 2017 13:35:02 +0000 (15:35 +0200)]
Remove ES3PlusWrapperContext
ES3PlusWrapperContext is not maintained anymore.
Components: Framework
Change-Id: I4327c219d6fd0362da89449884be48c1109a9b62
(cherry picked from commit
bda54aaefe2ce0c6bf05b1fc58b98721b9a9f1a2)
Ilia Mirkin [Sat, 9 Sep 2017 22:27:30 +0000 (18:27 -0400)]
Always apply flat qualifier to double inputs, same as int/uint
Affects:
KHR-GL45.enhanced_layouts.varying_structure_locations
Components: OpenGL
GitHub Issue: #53
(cherry picked from commit
878a6a5a8343b710ff7a7fcff0c219a25d0c4368)
Change-Id: I84de1d9b42695034a7b2f60c3fc5a19213733085
Piotr Byszewski [Tue, 19 Sep 2017 14:53:49 +0000 (16:53 +0200)]
Test EXT_shader_group_vote using variables
Current tests call shader group functions passing directly True/False
values. This change adds test that verifies new functions using
variables.
This change also reduces workgroup size for ES as all tests were
previously using workgroup size that was greater than the minimum
required.
Components: OpenGL
VK-GL-CTS issue: 693
Affects:
KHR-GL45.shader_group_vote.*
KHR-GLES31.core.shader_group_vote.*
Change-Id: Icbb5c278c65edd04fe339e31934df7624834c439
Jason Ekstrand [Fri, 6 Oct 2017 23:10:26 +0000 (16:10 -0700)]
VK_KHR_incremental_present: Trigger the fence off vkQueueSubmit
Previously, the tests were using AcquireNextImage to trigger the fence
and using it to guard the vkFreeCommandBuffers call. However, the
acquire fence tells you when the PE is done with an image and when you
can start rendering to it. This meant that the test was waiting for the
start of the previous frame's rendering not the end before deleting the
previous frame's command buffer. This meant that the previous frame's
rendering could still be active on the GPU when the command buffer was
freed. By triggering the fence off of the vkQueueSubmit, we wait for
the end of the previous frame's rendering before freeing its command
buffer.
VK-GL-CTS issue: 756
Component: Vulkan
Affects:
dEQP-VK.wsi.*.incremental_present.*
Change-Id: I87b25eb2a25e4ae48c8c527d5d63c45d89a31a30
(cherry picked from commit
db916576a76a17c02c14390fc5d5b3db1488d1d2)
Igor Ostrowski [Mon, 9 Oct 2017 14:41:31 +0000 (16:41 +0200)]
DrawUtil: Fix depthBounds check
Affects:
dEQP-VK.glsl.builtin_var.fragdepth*
Components: Vulkan
VK-GL-CTS issue: 758
Change-Id: I4da79b236921e64f11566fe386b144497249123b
Eleni Maria Stea [Wed, 20 Sep 2017 11:28:40 +0000 (14:28 +0300)]
Fix KHR-GL45/46.enhanced_layouts.varying_block_member_locations
The stages of the same program used in/out blocks that didn't match
(see section 4.7.1 in OpenGL 4.5 spec: qualification mismatch) and as a
result a linker error was produced. With this change all stages use the
same block definition which is either the 1st or the 2nd depending on
the iteration but not both at the same time:
Goku {
vec4 gohan;
vec4 goten;
vec4 chichi;
} gokuARRAY;
Goku {
layout (location = 2) vec4 gohan;
layout (location = 4) vec4 goten;
layout (location = 6) vec4 chichi;
} gokuARRAY;
Affects:
KHR-GL46.enhanced_layouts.varying_block_member_locations
KHR-GL45.enhanced_layouts.varying_block_member_locations
Components: OpenGL
VK-GL-CTS issue: 703
Change-Id: I03a6a90929e28489a2507838003c62475a8a58b8
(cherry picked from commit
4b6462e469359ac7ddec9c63d042e43ead6031fc)
Piotr Byszewski [Thu, 14 Sep 2017 14:06:49 +0000 (16:06 +0200)]
Increase negative GLSL coverage
GLSL and ESSL specify that the following are legal declarations:
int;
int ,a;
while the following are not:
struct foo { int; };
struct bar { int ,a; };
int a,;
Components: OpenGL
VK-GL-CTS issue: 7
Affects:
KHR-GL33.shaders.declarations.*
KHR-GLES3.shaders.declarations.*
Change-Id: Ib1735ca96e4ce7be60fa0ca91c64f7392b237605
Piotr Byszewski [Mon, 7 Aug 2017 12:12:55 +0000 (14:12 +0200)]
Increase coverage for robustness
Out-of-range fetching tests verify fetches 4K, 1MB and 10MB past
the end of the object.
Components: OpenGL
VK-GL-CTS issue: 16
Affects:
KHR-GL43.robust_buffer_access_behavior.*
KHR-GLES32.robust.robust_buffer_access_behavior.*
Change-Id: Ibad4eb3e30f3dc48b2d5ecb4781014a4374e4dc5
Piotr Byszewski [Fri, 15 Sep 2017 13:36:25 +0000 (15:36 +0200)]
Tests mixing UB/SSB with and without instance names
Matched uniform or shader storage block names (but not input or
output block names) must also either all be lacking an instance
name or all having an instance name.
Components: OpenGL
VK-GL-CTS issue: 3
Affects:
KHR-GL43.shader_storage_buffer_object.basic-name-match
KHR-GLES31.core.shader_storage_buffer_object.basic-name-match
KHR-GL33.shaders.uniform_block.common.name_matching
KHR-GLES3.shaders.uniform_block.common.name_matching
Change-Id: I7dc3476272d6f3fca3faa7e6bc57a008bce0e17f
Andrey Tuganov [Wed, 20 Sep 2017 20:42:35 +0000 (16:42 -0400)]
Fix non-unique type declarations in tests
Multiple test code generators were generating incorrect SPIR-V code
(non-unique type declarations of non-aggregate types).
Affects: dEQP-VK.spirv_assembly.*
Components: Vulkan
VK-GL-CTS issue: 195
Change-Id: I55200a2b78c84c298db05f34892aa4cffbfff74c
Jorg Wagner [Thu, 28 Sep 2017 11:59:51 +0000 (13:59 +0200)]
Correct image usage flags for YCbCr copy tests
The images are used as copy source and destination,
hence TRANSFER_SRC | TRANSFER_DST must be present
in the usage flags.
Images are not sampled from - remove SAMPLED flag.
Affects:
dEQP-VK.ycbcr.copy.*
Components: Vulkan
VK-GL-CTS issue: 715
Change-Id: Ie53d3e2fb18aa834af12938cd822aa6c50f75986
(cherry picked from commit
f2a5b768c81760bef727f1d72741e74274968518)
Panagiotis Apostolou [Thu, 7 Sep 2017 11:54:51 +0000 (13:54 +0200)]
Serialize region copies for YCbCr copy tests
As the generated regions can overlap - force all calls
of vkCmdCopyImage() to be executed sequentially by adding
a pipeline barrier between each call to avoid producing
undefined results.
Affects:
dEQP-VK.ycbcr.copy.*
Components: Vulkan
VK-GL-CTS issue: 716
Change-Id: Ifed67cdc077a8c03630b36f8aa59bd6dae0578a3
(cherry picked from commit
9fe7d4753721d10235cbf38194cf7c8cf2eb7d19)
Eleni Maria Stea [Wed, 20 Sep 2017 11:28:40 +0000 (14:28 +0300)]
Fix KHR-GL45/46.enhanced_layouts.varying_block_member_locations
The stages of the same program used in/out blocks that didn't match
(see section 4.7.1 in OpenGL 4.5 spec: qualification mismatch) and as a
result a linker error was produced. With this change all stages use the
same block definition which is either the 1st or the 2nd depending on
the iteration but not both at the same time:
Goku {
vec4 gohan;
vec4 goten;
vec4 chichi;
} gokuARRAY;
Goku {
layout (location = 2) vec4 gohan;
layout (location = 4) vec4 goten;
layout (location = 6) vec4 chichi;
} gokuARRAY;
Affects:
KHR-GL46.enhanced_layouts.varying_block_member_locations
KHR-GL45.enhanced_layouts.varying_block_member_locations
Components: OpenGL
VK-GL-CTS issue: 703
Change-Id: I03a6a90929e28489a2507838003c62475a8a58b8
Piotr Byszewski [Mon, 11 Sep 2017 16:31:16 +0000 (18:31 +0200)]
Test struct names hiding other struct names
Test if names are correctly hidden when a level of nesting is
involved.
Components: OpenGL
VK-GL-CTS issue: 14
Affects:
KHR-GLES3.shaders.name_hiding.*
Change-Id: I3eb83bfcf6d10fcb15abaa6c086ecd5f9c3d0b36
Iago Toral Quiroga [Fri, 22 Sep 2017 09:06:50 +0000 (11:06 +0200)]
Fix locations in enhanced layouts tests
Some of the tests query the last output location in
each shader stage and write an output to them that is
then read in the next shader stage, but they do not
consider if that location exists in the next shader
stage.
Fix this by selecting the maximum output location to
use as the minimum between the stage's maximum output
location and the next stage's maximum input location.
The same issue applies to the maximum input location
available in each shader stage.
Components: OpenGL
VK-GL-CTS issue: 708
Affects:
KHR-GL45.enhanced_layouts.varying_components
KHR-GL45.enhanced_layouts.varying_locations
KHR-GL45.enhanced_layouts.varying_location_limit
KHR-GL46.enhanced_layouts.varying_components
KHR-GL46.enhanced_layouts.varying_locations
KHR-GL46.enhanced_layouts.varying_location_limit
Change-Id: Iacf84bd3eeb33eadaf5a9a816f7ebb7205492af5
Ari Suonpaa [Thu, 21 Sep 2017 06:13:33 +0000 (09:13 +0300)]
Fixed invalid OpPtrAccessChain usage in access chain tests.
Modified dEQP-VK.spirv_assembly.instruction.*.indexing.* tests to
use an array of structs as input to have a meaningful usage of
OpPtrAccessChain.
Affects:
dEQP-VK.spirv_assembly.instruction.*.indexing.*
Components: Vulkan
VK-GL-CTS issue 686
Change-Id: Ia1604d5a24712c522f51cbd0692d85b1baa2d27c
Piotr Byszewski [Wed, 13 Sep 2017 07:56:42 +0000 (09:56 +0200)]
Test BLOCK_INDEX of uniform inside block array
Uniform blocks declared in an array are considered active
if any member of the array would otherwise be considered
active.
Components: OpenGL
VK-GL-CTS issue: 13
Affects:
KHR-GL43.program_interface_query.uniform-block-array
KHR-GLES31.core.program_interface_query.uniform-block-array
Change-Id: Iaa67be6f23f168e35e5ddc3965342756dc2c59ed
Nicolai Hähnle [Wed, 20 Sep 2017 20:26:44 +0000 (22:26 +0200)]
Allow min/max/clamp to flush subnormals only during comparison
An implementation may follow the rule that subnormals are flushed
to zero for all arithmetic operations (including comparisons), but
never for mere copies.
In this case, if the input to min/max/clamp contains more than one
subnormal number (including +/-0.0), the returned value may be any
of the subnormal numbers and not really the true minimum/maximum/
clamped value.
This does not hurt real applications, since any subsequent use of
the result will flush it to zero anyway, and it does seem to be
implicitly allowed by the GLSL ES spec.
Change the implementation of min/max/clamp to explicitly handle
the subnormal case in the comparison.
Components: AOSP
VK-GL-CTS issue: 705
Affects:
dEQP-GLES3.functional.shaders.builtin_functions.precision.clamp.*
dEQP-GLES3.functional.shaders.builtin_functions.precision.max.*
dEQP-GLES3.functional.shaders.builtin_functions.precision.min.*
dEQP-GLES31.functional.shaders.builtin_functions.precision.clamp.*
dEQP-GLES31.functional.shaders.builtin_functions.precision.max.*
dEQP-GLES31.functional.shaders.builtin_functions.precision.min.*
Change-Id: I9a5bed9d78e311f96fe03b689299d777050ea063
Adam Czupryna [Thu, 29 Jun 2017 14:50:53 +0000 (16:50 +0200)]
Add CTS_ARB_gl_spirv test specification proposal
This is CTS_ARB_gl_spirv test specification proposal.
Components: OpenGL
VK-GL-CTS issue: 554
Change-Id: I95478228f66c27ed5ff78a9e17a0381512abbf21
Paavo Pessi [Thu, 21 Sep 2017 10:08:13 +0000 (13:08 +0300)]
Test vkCmdBlitImage for cubemaps with layerCount=6
Tests were added to verify that the typical use-case of generating
mipmap levels for cubemaps with vkCmdBlitImage() works with all supported
formats.
Tests in 'from_base_level' group use the base image as a source for all
mip levels and generate all mip levels for all six image layers with a
single blit command.
Tests in 'from_previous_level' group use the previous mip level as a
source for the next mip level and generate each mip level for all
six image layers with a separate blit command with barriers between
each blit.
Existing mipmap tests cases in the
dEQP-VK.api.copy_and_blit.*.blit_image.all_formats.generate_mipmaps.*
test group were moved to
dEQP-VK.api.copy_and_blit.*.blit_image.all_formats.generate_mipmaps.*.layercount_1.*
New tests:
dEQP-VK.api.copy_and_blit.core.blit_image.all_formats.generate_mipmaps.*.layercount_6.*
dEQP-VK.api.copy_and_blit.dedicated_allocation.blit_image.all_formats.generate_mipmaps.*.layercount_6.*
Components: Vulkan
VK-GL-CTS issue: 615
VK-GL-CTS public issue: 50
Change-Id: I512484434dda155b668a600c8303650d42417e0a
Alexander Galazin [Thu, 28 Sep 2017 10:55:33 +0000 (12:55 +0200)]
Merge vk-gl-cts/vulkan-cts-1.0.2 into vk-gl-cts/master
Change-Id: Ic02c027d4172d581d782eec10ba3bf8955c2c356
Paavo Pessi [Fri, 22 Sep 2017 09:00:16 +0000 (12:00 +0300)]
Validate empty render pass
The test creates a render pass and a subpass with no input, color,
resolve or depth/stencil attachments and executes a single draw
command.
New tests:
dEQP-VK.renderpass.suballocation.simple.no_attachments
dEQP-VK.renderpass.dedicated_allocation.simple.no_attachments
Components: Vulkan
VK-GL-CTS issue: 620
VK-GL-CTS public issue: 10
Change-Id: I0da8a6850898979e65d39f323b14694968d9d6c5
David Neto [Thu, 14 Sep 2017 22:03:37 +0000 (18:03 -0400)]
Test reading compute shader builtin vars by component
- Adds a test of reading compute shader builtin variables by component.
These get test case name suffix of "_component".
- Previous behaviour, reading the whole builtin variable, is preserved
in cases but without an extra suffix.
- WorkgroupSize builtin is made specializable to prevent Glslang
from constant folding its value.
Affects:
Components: Vulkan
dEQP-VK.compute.builtin_var.*
VK-GL-CTS issue: 694
Change-Id: Ia783babf4ba5e24fc58bc9d19594bb7d921de7c9
Alexander Galazin [Wed, 27 Sep 2017 12:13:30 +0000 (14:13 +0200)]
Remove YCbCr from the instance extensions
VK_KHR_sampler_ycbcr_conversion is a device extension.
It must not be listed as an allowed instance extension.
Components: Vulkan
VK-GL-CTS issue: 698
Affects:
dEQP-VK.api.info.instance.extensions
Change-Id: I1ee1448535df75016a0214ebfe800a476daf32c4
Piotr Byszewski [Tue, 19 Sep 2017 11:58:53 +0000 (13:58 +0200)]
Fix UniformBlockPrecisionMatching test
Fixed an issue that resulted in omiting second set of
shaders (first set was tested twice).
Components: OpenGL
Affects:
KHR-GLES3.shaders.uniform_block.common.precision_matching
Change-Id: I44205f66cd5672692fc46f4db91801b86aa8ff74