Alexander Galazin [Wed, 25 Mar 2020 07:46:07 +0000 (08:46 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I36c436cd72ec705b74e162b7bd11aff4a684f123
Graeme Leese [Fri, 20 Mar 2020 13:30:31 +0000 (13:30 +0000)]
More vertex-invocation fixes for synchronisation tests
The fix for multiple vertex-shader instances was missed when the
timeline semaphore version of these tests was added. This is the fix
from https://gerrit.khronos.org/c/4451 applied to these new tests.
Components: Vulkan
VK-GL-CTS issue: 2253, 1958
Affects: dEQP-VK.synchronization.op.*.timeline_semaphore.write_indirect_buffer_draw*
Change-Id: I6013ef88a0acd362948c179416928c1d3fc064dd
Alexander Galazin [Wed, 4 Mar 2020 10:06:51 +0000 (11:06 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: Ifeab9cfd06891288494e9080d40819c62a2092a7
Alexander Galazin [Wed, 4 Mar 2020 07:50:07 +0000 (08:50 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.4 into vk-gl-cts/vulkan-cts-1.1.5
Change-Id: I6221b31126a23415f4aceb05f91fc8c7b166c876
Alexander Galazin [Thu, 27 Feb 2020 11:20:44 +0000 (12:20 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.3 into vk-gl-cts/vulkan-cts-1.1.4
Change-Id: Ie332664c1049e2e1339f60aa7dd2615d1aa83105
dzabaj01 [Mon, 24 Feb 2020 12:39:21 +0000 (13:39 +0100)]
Fix dEQP-VK.geometry.layered.2d_array.readback
The patch fixes several image memory barriers that
are missing the (color or depth/stencil) read access
mask. The read access mask is needed because the
attachments are setup with the VK_ATTACHMENT_LOAD_OP_LOAD
load operation.
Components: Vulkan
VK-GL-CTS Issue: 2227
Affects: dEQP-VK.geometry.layered.2d_array.readback
Change-Id: I22b696f449cb83d88f3ec15d928577d6b3d65b6f
James Fitzpatrick [Mon, 20 Jan 2020 09:02:24 +0000 (09:02 +0000)]
Correctly handle normalized formats in FloatFormat
For the texture explict lod tests normalized values were treated as 8
bit floating point formats when calculating the reference interval.
This ignores the fact that normalized formats are encoded as 1 / N,
where N is an integer, this results in the rounding used in tests not
being able to accurately round normalized formats.
This incorrect rounding causes the acceptable ranges to be
significantly smaller than other formats and extremely asymetric
around the ideal filtered value for certain samples.
Affects: dEQP-VK.texture.explicit_lod.*
Components: Vulkan
VK-GL-CTS issue: 2189
Change-Id: Ia4f3adefab4bfb5d9c91bd7033199b0e107179d5
Alexander Galazin [Thu, 20 Feb 2020 14:41:31 +0000 (15:41 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: If35028985c57f66a6fc3f3d11ac1a9485493f7f2
James Fitzpatrick [Tue, 11 Feb 2020 11:58:04 +0000 (11:58 +0000)]
Return Quality Warning for inaccurate filtering
For cases when the relaxed precision has to be used to verify a sample
in the explicit lod tests return a quality warning rather than a pass.
Affects: dEQP-VK.texture.explicit_lod.*
Components: Vulkan
VK-GL-CTS issue: 2189
Change-Id: If1cad3ae74bc91636527215f6871faa2e1feee79
dzabaj01 [Tue, 18 Feb 2020 12:17:18 +0000 (13:17 +0100)]
Fix dEQP-VK.geometry.layered.2d_array.readback
The patch fixes several image memory barriers that
are missing the (color or depth/stencil) read access
mask. The read access mask is needed because the
attachments are setup with the VK_ATTACHMENT_LOAD_OP_LOAD
load operation.
Components: Vulkan
VK-GL-CTS Issue: 2227
Affects: dEQP-VK.geometry.layered.2d_array.readback
Change-Id: Ie96c93e6d8fee1ea0c0a63417707ec7479551413
Martin Freebody [Wed, 20 Nov 2019 10:49:25 +0000 (10:49 +0000)]
Protected Memory allocation tests require compatible device
For memory allocation tests that detect protected memory capability,
ensure that the device being used first enables the protected memory
feature. If the physical device does not support protected memory,
the allocation type is skipped in the test.
Affects: dEQP-VK.memory.allocation.*
Components: Vulkan
VK-GL-CTS issue: 2163
Change-Id: I899c7cf7e13d04f2674de47c3bb4d08f3de589f4
Martin Freebody [Wed, 20 Nov 2019 10:49:25 +0000 (10:49 +0000)]
Protected Memory allocation tests require compatible device
For memory allocation tests that detect protected memory capability,
ensure that the device being used first enables the protected memory
feature. If the physical device does not support protected memory,
the allocation type is skipped in the test.
Affects: dEQP-VK.memory.allocation.*
Components: Vulkan
VK-GL-CTS issue: 2163
Change-Id: I899c7cf7e13d04f2674de47c3bb4d08f3de589f4
Alexander Galazin [Thu, 13 Feb 2020 12:16:07 +0000 (13:16 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I0b304876f100815c74515584179d954ac0f65788
John Anthony [Wed, 29 Jan 2020 15:33:53 +0000 (16:33 +0100)]
Only require geometryStreams feature in XFB query tests if needed
Change transform feedback query tests to only require the
geometryStreams feature for test cases using stream ID != 0.
Affects:
- dEQP-VK.transform_feedback.simple.query_0_*
- dEQP-VK.transform_feedback.simple.host_query_reset_0_*
Components: Vulkan
VK-GL-CTS Issue: 2197
Change-Id: Id1378f58e6634dd70ad89843e000d047af0ea496
Mikel Garai [Mon, 20 Jan 2020 08:30:53 +0000 (09:30 +0100)]
Fix transform feedback triangle strip verification
Vulkan specifies that the first vertex in a multi vertex primitive is
implementation specific, but the winding must be respected.
This change allows any first vertex as long as the vertices and winding
are correct.
Components: Vulkan
Affects:
dEQP-VK.transform_feedback.simple.triangle_strip_with_adjacency_*
VK-GL-CTS Issue: 2193
Change-Id: I3d195c50a4b3b1f51529014fb97cebef19f872b7
Tom Cooper [Mon, 3 Feb 2020 14:39:33 +0000 (14:39 +0000)]
Fix dEQP-VK.api.image_clearing.*.clear_depth_stencil_image
Check device extensions before trying to create the renderpass using
vkCreateRenderPass2KHR.
Components: Vulkan
VK-GL-CTS Issue: 2208
Affects: dEQP-VK.api.image_clearing.*
Change-Id: I02bb1ccb8f07c7f2baa7357a639f4bdcc1dcd73f
Szymon Szczyrbak [Tue, 14 Jan 2020 12:40:51 +0000 (13:40 +0100)]
Add memory barrier to fragment shader interlock tests.
Image or buffer written by GPU are copied to host-visible buffer used to
verify test's results.
Added a TRANSFER_WRITE/HOST_READ pipeline barrier to make sure that the data
accessed by host is up-to-date.
Component: Vulkan
Affects: dEQP-VK.fragment_shader_interlock.basic.*
VK-GL-CTS Issue: 2165
Change-Id: I6d458b2e650a3813e62948d670b8a53e864aed37
Juan A. Suarez Romero [Wed, 8 Jan 2020 11:58:40 +0000 (11:58 +0000)]
Modify test line position in LinesTestInstance::generateLines()
This slightly change a test line position in order to fix
dEQP-VK.rasterization.primitives.no_stipple.smooth_lines_wide, to avoid
a false failure according to comments in CL#5152.
Affects: dEQP-VK.rasterization.primitives.*
Components: Vulkan
Change-Id: Ie8bde1d1d32f76939ce4a4ba21aaec4c471edd63
Alexander Galazin [Mon, 13 Jan 2020 14:34:33 +0000 (15:34 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I654934956a489ddcdec47ac8fc3e2fe638929950
Alexander Galazin [Mon, 13 Jan 2020 12:10:10 +0000 (13:10 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.4 into vk-gl-cts/vulkan-cts-1.1.5
Change-Id: I420bc4aa41a2f7a67434b4f674c3d3aa161dc6c3
Alexander Galazin [Mon, 13 Jan 2020 10:54:20 +0000 (11:54 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.3 into vk-gl-cts/vulkan-cts-1.1.4
Change-Id: Id9ed312f96fa266a0e3809211b14566ece03e791
Alexander Galazin [Mon, 13 Jan 2020 09:45:56 +0000 (10:45 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.2 into vk-gl-cts/vulkan-cts-1.1.3
Change-Id: I37d5b18d5980372057827bc4435239be98fd30bc
Alan Baker [Mon, 25 Nov 2019 18:03:36 +0000 (13:03 -0500)]
Fix SPIR-V assembly tests with bad cfgs
* Update SPIR-V assembly tests to not declare a continue target as a
merge block
Component: Vulkan
VK-GL-CTS Issue: 2122
Affects:
dEQP-VK.spirv_assembly.instruction.graphics.switch_block_order.out_of_order*
dEQP-VK.spirv_assembly.instruction.graphics.opphi.out_of_order*
dEQP-VK.spirv_assembly.instruction.graphics.loop.multi_block_loop_construct*
dEQP-VK.spirv_assembly.instruction.graphics.loop.multi_block_continue_construct*
dEQP-VK.spirv_assembly.instruction.graphics.loop.continue*
dEQP-VK.spirv_assembly.instruction.graphics.loop.break*
dEQP-VK.spirv_assembly.instruction.graphics.loop.return*
Change-Id: I370331414aae27b53ed79b448b72581e2352c3d7
Alexander Galazin [Sun, 5 Jan 2020 08:33:44 +0000 (11:33 +0300)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: Id7608050ab3f1c5ce3d26a1480af0de7d90d75ff
Alexander Galazin [Sun, 5 Jan 2020 07:48:47 +0000 (10:48 +0300)]
Merge vk-gl-cts/vulkan-cts-1.1.4 into vk-gl-cts/vulkan-cts-1.1.5
Change-Id: I51213fc50189215a04c9206a8b805962052235cb
Alexander Galazin [Sun, 5 Jan 2020 06:36:56 +0000 (09:36 +0300)]
Merge vk-gl-cts/vulkan-cts-1.1.3 into vk-gl-cts/vulkan-cts-1.1.4
Change-Id: I2a051b8a36b997c3f904ddd78657de8c1719a53b
Matthew Netsch [Fri, 20 Dec 2019 20:46:11 +0000 (15:46 -0500)]
Allows NaNs in an assert
Components: Vulkan
VK-GL-CTS Issue: 2103
Affects:
dEQP-VK.spirv_assembly.instruction.compute.float16.arithmetic_*
Change-Id: I5848660d120b5a4a07df589a74f47b8ca30befb5
Matthew Netsch [Fri, 20 Dec 2019 20:43:04 +0000 (15:43 -0500)]
Adds subgroup ops stage check
Components: Vulkan
VK-GL-CTS Issue: 2084
Affects:
dEQP-VK.memory_model.message_passing.*.subgroup.*
dEQP-VK.memory_model.write_after_read.ext.*.subgroup.*
Change-Id: Ibb820b4d3de1782bc844f3eb32b6e752745e9976
Alexander Galazin [Sat, 21 Dec 2019 09:50:50 +0000 (10:50 +0100)]
Re-enable tests
Change-Id: Ic4d90d99b69ef931ba2f3f49538201762eb6746e
Matthew Netsch [Fri, 4 Oct 2019 15:46:04 +0000 (11:46 -0400)]
Account for VS executing more than once when atomics are used.
Components: Vulkan
Affects: dEQP-VK.glsl.opaque_type_indexing.atomic_counter.*
dEQP-VK.glsl.atomic_operations.*
Change-Id: I15eabf9b47af204fd13370efdc323af141b6b11b
Alexander Galazin [Fri, 20 Dec 2019 12:50:25 +0000 (13:50 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I0f9446b635621472d72566a6176c7d0c00306263
Dae Kim [Fri, 13 Dec 2019 09:26:05 +0000 (09:26 +0000)]
Avoid undefined behaviour when calculating subgroup ballot masks
The affected tests are shifting the bits of the last component of the
computed ballot masks by 32. However, the SPIR-V specification
establishes that the result of OpShift* is undefined if `Shift` is
greater than *or equal* to the bit width of the components of `Base`.
Affects:
dEQP-VK.subgroups.ballot_other.*
Components: Vulkan
VK-GL-CTS issue: 2083
Change-Id: I8c9b4373f1429669a40741986916e8d71b7eb573
Piers Daniell [Wed, 11 Dec 2019 22:19:18 +0000 (15:19 -0700)]
Fix bad merge from https://gerrit.khronos.org/#/c/5107/
When https://gerrit.khronos.org/#/c/4874/ was cherry-picked
to the vulkan-cts-1.1.6 branch it would have hit conflicts
with the new VK_EXT_line_rasterization tests from
https://gerrit.khronos.org/#/c/3997/.
This CL fixes a bad conflict resolution in
tcuRasterizationVerifier.cpp that forced the verification
mode of all line types top be scrict, when they should
have got their verification mode from
scene.verificationMode.
This CL also modifies the position of the test lines
for case 2 of LinesTestInstance::generateLines(). The
position of the last line, added with CL 5107, was moved
lower to avoid excessive overlap with the other three
lines. This excessive overlap caused many pixels to
share edges and aggravate a minor flaw in the software
line rasterizer where it always assumes full coverage on
pixels that share edges of two or more triangles.
They may actually still be only partial coverage when
doing smooth or multisample rendering, and forcing it
to full coverage won't match what hardware does.
Affects: dEQP-VK.rasterization.primitives.*
Components: Vulkan
VK-GL-CTS Issue: 2062
Change-Id: Ie865f46c24e8265fd47354faaf9be5efd002b2d4
Alexander Galazin [Tue, 17 Dec 2019 15:30:21 +0000 (16:30 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I7054b3a97f5f57fb5d5829f21991efe70108c77a
Alexander Galazin [Tue, 17 Dec 2019 14:34:21 +0000 (15:34 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.4 into vk-gl-cts/vulkan-cts-1.1.5
Change-Id: I4d9b046396a757d9fce9a0a326b150deb37015ce
Alexander Galazin [Tue, 17 Dec 2019 12:32:46 +0000 (13:32 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.3 into vk-gl-cts/vulkan-cts-1.1.4
Change-Id: Ia01386a2e8fec477414b8488a7bfa883265c32ec
Alexander Galazin [Tue, 17 Dec 2019 10:58:46 +0000 (11:58 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.2 into vk-gl-cts/vulkan-cts-1.1.3
Change-Id: Iecc5fe7d8f5119a46a78fd7d6f3c3b613c45059d
Paul Thomson [Mon, 9 Dec 2019 16:03:38 +0000 (16:03 +0000)]
Re-include some Android CTS tests
Components: Android
Included tests:
dEQP-VK.glsl.builtin.precision.*
dEQP-VK.spirv_assembly.instruction.compute.opfunord.*
dEQP-VK.spirv_assembly.instruction.*.indexing.*u32*
dEQP-VK.spirv_assembly.instruction.*.indexing.*s32*
Change-Id: Ib9be1f9ef388c7baeaf8f11b1dd7e60efdf72dc5
Paul Thomson [Thu, 5 Dec 2019 18:13:00 +0000 (18:13 +0000)]
Provide a highp variable to MAKE_HIGH_BALLOT_RESULT
The MAKE_HIGH_BALLOT_RESULT macro assumes the input is 32bit.
Given a mediump variable, ex: gl_SubgroupSize, would produce an
undefined result when the input value is >= 16 and the computation
is in 16 bits.
Provide a highp input would ensure the computation is in 32 bits.
Affects:
dEQP-VK.subgroups.ballot_other.*
Components: vulkan
VK-GL-CTS issue: 1894
Change-Id: I22987f20be3d08444c52cb722fc23494777a5636
Victor Khimenko [Tue, 10 Sep 2019 21:25:00 +0000 (23:25 +0200)]
Fix incorrect use of dlopen/dlclose
Function pointers returned by dlsym are retained after dlclose() is
called. If the library is actually unloaded by that dlclose() call, then
calling through those function pointers then triggers undefined
behavior.
Components: Vulkan
Affects: dEQP-VK.api.external.memory.android_hardware_buffer.*
Google Bug: b/
140826027
Change-Id: I1da2d10e97e5567578bf06d768820764882f7126
Arkadiusz Sarwa [Fri, 15 Nov 2019 16:35:24 +0000 (17:35 +0100)]
Fix memory leaks in dEQP-VK.api.invariance.random
Test dEQP-VK.api.invariance.random contains memory leaks.
Change fixes memory leaks.
Components: Vulkan
VK-GL-CTS Issue: 2110
Affects: dEQP-VK.api.invariance.random
Change-Id: I4ccd9d87b69f51da661ca85f10eb347086594aaa
Alastair Donaldson [Tue, 26 Nov 2019 14:17:11 +0000 (14:17 +0000)]
Fix invalid SPIR-V in GraphicsFuzz tests
Components: Vulkan
VK-GL-CTS Issue: 2122
Affected tests:
dEQP-VK.graphicsfuzz.continue-and-merge
dEQP-VK.graphicsfuzz.control-flow-switch
dEQP-VK.graphicsfuzz.discard-continue-return
dEQP-VK.graphicsfuzz.fragcoord-control-flow-2
dEQP-VK.graphicsfuzz.fragcoord-control-flow
dEQP-VK.graphicsfuzz.loop-call-discard
dEQP-VK.graphicsfuzz.mat-array-deep-control-flow
dEQP-VK.graphicsfuzz.mat-array-distance
dEQP-VK.graphicsfuzz.return-in-loop-in-function
dEQP-VK.graphicsfuzz.unreachable-continue-statement
Change-Id: Iaa7448ee4a815d9e3a1c8a5582214ecd599ac143
Alexander Galazin [Thu, 5 Dec 2019 10:48:22 +0000 (11:48 +0100)]
Whitelist Vulkan CTS 1.1.6.3
Affects: dEQP-VK.api.driver_properties.properties
Components: Vulkan
Change-Id: Ib46d2ea03898930641539bd0ad1cf657c2994aa4
Jeff Bolz [Tue, 24 Sep 2019 22:01:10 +0000 (17:01 -0500)]
uvec2 tests for VK_KHR_buffer_device_address
Add variants of the 'convert' tests that convert to and from uvec2
rather than uint64_t.
Components: Vulkan
VK-GL-CTS Issue: 2018
Affects: dEQP-VK.binding_model.buffer_device_address.*
Change-Id: I9e1a2f031e295b9200faff571aae2fc5cbb9005c
(cherry picked from commit
bcc3bf2dee4bcd8505c1b0c16ee05cafad185749)
Jeff Bolz [Tue, 9 Jul 2019 04:09:44 +0000 (23:09 -0500)]
Tests for VK_KHR_buffer_device_address
Components: Vulkan
Affects: dEQP-VK.binding_model.buffer_device_address*
Affects: dEQP-VK.compute.cooperative_matrix*phys*
Affects: dEQP-VK.memory_model.*phys*
Affects: dEQP-VK.ssbo.*phys*
Affects: dEQP-VK.spirv_assembly.*phys*
Change-Id: I045b5832e9857f6e826957e204237a32e0bcbce4
(cherry picked from commit
86db0cf343430c2b434f9313c6c4ef5188d9bc23)
Ben Clayton [Wed, 4 Dec 2019 17:13:41 +0000 (17:13 +0000)]
Bump timeout for the basic fence tests
10ms for the LONG_FENCE_WAIT could cause spurious test failures for CPU-based
Vulkan implementations when the CPU is under heavy load.
This is now bumped to a second, but assuming the test passes, this should never
block for the entire second.
Affects: dEQP-VK.synchronization.basic.fence.*
Components: Vulkan
Change-Id: Ice07a1697b57e789b9f57f4a5da663c83a0c132a
Alexander Galazin [Thu, 5 Dec 2019 16:22:40 +0000 (17:22 +0100)]
Temporarily remove lines_wide tests
Components: Vulkan
Affects:
dEQP-VK.rasterization.primitives.no_stipple.smooth_lines_wide
dEQP-VK.rasterization.primitives_multisample_2_bit.no_stipple.smooth_lines_wide
dEQP-VK.rasterization.primitives_multisample_4_bit.no_stipple.smooth_lines_wide
dEQP-VK.rasterization.primitives_multisample_8_bit.no_stipple.rectangular_line_strip_wide
dEQP-VK.rasterization.primitives_multisample_8_bit.no_stipple.smooth_lines_wide
dEQP-VK.rasterization.primitives_multisample_16_bit.no_stipple.smooth_lines_wide
Change-Id: Id74c00f56519d03fb300a4a06bc20f8f0365bd3e
Graeme Leese [Thu, 5 Dec 2019 10:19:01 +0000 (10:19 +0000)]
Fix RGB9E5 encode
The overflow correction was broken and resulted in encoded values being
2x too large.
Components: Framework
VK-GL-CTS issue: 2132
Affects: dEQP-VK.*e5*
Change-Id: Ibcab4c889b044c8668aae3c7a1288b8cda1ff0e6
Alexander Galazin [Thu, 5 Dec 2019 09:37:17 +0000 (10:37 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: Ie910031c0296648e7ff34fb69cdbed0de5f24a1b
Alexander Galazin [Thu, 5 Dec 2019 08:38:48 +0000 (09:38 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.4 into vk-gl-cts/vulkan-cts-1.1.5
Change-Id: I13e1238c4bfc3cb1366e25def3b081591a8101de
Slawomir Cygan [Fri, 29 Nov 2019 13:06:03 +0000 (14:06 +0100)]
Do the 'from UNDEFINED' layout transition only on the first pass
The test uses multiple submissions, each rendering to a part
of the same render target. Only the first submission should do
the transition 'from' UNDEFINED layout, otherwise each pass (submit)
would invalidate render target contents from previous submits.
Components: Vulkan
VK-GL-CTS Issue: 2131
Change-Id: Ic424e391e0348f6333b50d4c0f4f57d161ec95fb
Affects: dEQP-VK.descriptor_indexing.*
Alexander Galazin [Thu, 5 Dec 2019 07:47:58 +0000 (08:47 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.3 into vk-gl-cts/vulkan-cts-1.1.4
Change-Id: I4d9bef08ce50fbba4d6a61cbba1e3cdb28051543
Yanjun Zhang [Sat, 16 Nov 2019 05:54:57 +0000 (21:54 -0800)]
Add two more formats with proper threshold values.
Need to add two more formats with proper threshold values in
vkApiCopiesAndBlittingTests.cpp: getFormatThreshold() to make
CTS pass on some VSI mobile GPUs.
Affects:
dEQP-VK.api.copy_and_blit.core.blit_image.simple_tests.*
dEQP-VK.api.copy_and_blit.core.blit_image.all_formats.*
dEQP-VK.api.copy_and_blit.dedicated_allocation.blit_image.*
Components: Vulkan
VK-GL-CTS issue: 2112
Change-Id: Ia3a314fd825ed13c44e49720fffffa2dff8f26af
Alexander Galazin [Tue, 3 Dec 2019 18:46:18 +0000 (19:46 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I0845a703755ac572d54f56394a04a9a4ace2fdce
Mike Byrne [Mon, 13 May 2019 14:15:02 +0000 (15:15 +0100)]
Extend rasterization testing, modify verification
dEQP-VK.rasterization.primitives.lines tests are extended to add tests
drawing vertical and horizontal lines.
Extend dEQP-VK.rasterization.primitives.lines and
dEQP-VK.rasterization.primitives_multisample_n_bit.lines to test at 258
resolution (a non power of two resolution).
The strict line verification coverage is modified to allow checking of
vertical and horizontal lines. Previously pixels in vertical or hori-
zontal lines were marked as partially covered and hence the test did
not detect whether the lines were rendered correctly or not.
Affects: dEQP-VK.rasterization.primitives*
Components: Vulkan
VK-GL-CTS Issue: 2062
Change-Id: Ifb76a44ba1ebdc97353bbd1a4b2552e415f2723d
Samuel Iglesias Gonsálvez [Wed, 27 Nov 2019 10:06:54 +0000 (11:06 +0100)]
Check memory model support in volatile atomic tests
Volatile atomic tests were not checking vulkan memory model support.
Components: Vulkan
VK-GL-CTS issue: 2125
Affected tests:
dEQP-VK.spirv_assembly.*atomic*
Change-Id: I67557361e2df9fc057ab38a1104f66512a532cf0
(cherry picked from commit
5ded1abda3169e29a6116b6e6b53ce0b9a6d5ab2)
Matthew Netsch [Fri, 15 Nov 2019 17:29:30 +0000 (12:29 -0500)]
Add feature checks to subgroups tests
Components: Vulkan
VK-GL-CTS Issue: 2082
Affects:
dEQP-VK.subgroups.*
Change-Id: Ide9aeaf106dcf1efbc2cf90cd6b46a61586c4f72
Mike Byrne [Mon, 13 May 2019 14:15:02 +0000 (15:15 +0100)]
Extend rasterization testing, modify verification
dEQP-VK.rasterization.primitives.lines tests are extended to add tests
drawing vertical and horizontal lines.
Extend dEQP-VK.rasterization.primitives.lines and
dEQP-VK.rasterization.primitives_multisample_n_bit.lines to test at 258
resolution (a non power of two resolution).
The strict line verification coverage is modified to allow checking of
vertical and horizontal lines. Previously pixels in vertical or hori-
zontal lines were marked as partially covered and hence the test did
not detect whether the lines were rendered correctly or not.
Affects: dEQP-VK.rasterization.primitives*
Components: Vulkan
VK-GL-CTS Issue: 2062
Change-Id: Ifb76a44ba1ebdc97353bbd1a4b2552e415f2723d
Piers Daniell [Wed, 13 Nov 2019 21:01:10 +0000 (14:01 -0700)]
Fix SFR swapchain test's implicit layout transitions
The Vulkan spec says that layout transitions done on peer bound
image memory requires the implementation to support
VK_PEER_MEMORY_FEATURE_GENERIC_SRC_BIT and
VK_PEER_MEMORY_FEATURE_GENERIC_DST_BIT. The test never checked
for this capability and did implicit render pass layout
transitions anyway.
This CL updates the test to use an explicit layout transitions
on local aliases of the image when the implementation doesn't
support doing peer-to-peer transfer.
This CL also fixes a bug in the SFR rects definition in
VkBindImageMemoryDeviceGroupInfo to specify the correct number
of rectangles.
Affects:
dEQP-VK.wsi.win32.swapchain.render.*
dEQP-VK.wsi.win32.swapchain.modify.*
Components: Vulkan
VK-GL-CTS issue: 2094
Change-Id: I31b7099a6d5c4f11be83d6d14a5c02cca6408441
John Corbally [Thu, 7 Nov 2019 16:01:18 +0000 (16:01 +0000)]
Apply the same relaxed precision fix for interface_blocks.* to basic_type.*
Fix relaxed precision and no perspective dEQP tests
In the RelaxedPrecision case, change epsilon to 2e-3 for a 16-bit float.
In the NoPerspective case, change epsilon to 3e-7 for a 32-bit float.
The epsilon is now relative to the correct value.
Affects:
dEQP-VK.spirv_assembly.instruction.graphics.cross_stage.basic_type.*
Components: Vulkan
VK-GL-CTS issue: 1818
Change-Id: I615e65f2f6e4287dc41ff981f18972cf14bcb374
Slawomir Cygan [Thu, 21 Nov 2019 16:35:23 +0000 (17:35 +0100)]
Remove verification of string bytes after null termination.
Spec does not require to zero bytes after the null-termination of the string.
Components: Vulkan
VK-GL-CTS Issue: 2117
Affects: dEQP-VK.pipeline.executable_properties.*
Change-Id: Ia07b51f22379cea28ca73fe0a1594ef5230c27af
Piotr Byszewski [Tue, 19 Nov 2019 13:48:46 +0000 (14:48 +0100)]
Fix cbuffer packing test
Components: Vulkan
VK-GL-CTS issue: 2109
Affects:
dEQP-VK.spirv_assembly.instruction.compute.hlsl_cases.cbuffer_packing
Change-Id: Iebedd63004bceed7542a433da9be3d43a59fed2e
Alexander Galazin [Wed, 20 Nov 2019 08:57:05 +0000 (09:57 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I6690ecfeeaf879819d23bb2a8a65c1713809e6a1
Juasheem Sultan [Fri, 1 Nov 2019 22:26:14 +0000 (15:26 -0700)]
Re-align importing dedicated memory to spec
Re-implemented importing dedicated memory for synchronization test to
better align with the Vulkan spec. The current solution
makes the assumption that not requiring memory to be dedicated upon
import implies that the memory is required to not be dedicated, and this
does not appear to align with the spec.
Affects:
* dEQP-VK.synchronization.signal_order.*
Components: Vulkan
VK-GL-CTS issue: 2071
Change-Id: Ie1ada8e4966e044ff0d265dce65f4a0680b311aa
Alexander Galazin [Mon, 18 Nov 2019 12:39:00 +0000 (13:39 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I163fd1d3c3f3a600e3a89031d350eb2b08f81e4d
Chris Mayer [Fri, 8 Nov 2019 12:28:16 +0000 (13:28 +0100)]
Add memory barrier to VK transform_feedback tests
The transform feedback tests are reading the transform
feedback buffers which was written by the GPU from the
host. Added a memory barrier to make sure the data is up to
date when it is accessed by the host.
Affects: dEQP-VK.transform_feedback.*
Components: Vulkan
VK-GL-CTS issue: 2096
Change-Id: Ibf9f86e9c1f02bdbfe822f62c44c30155ed8c555
Alexander Galazin [Mon, 18 Nov 2019 11:28:51 +0000 (12:28 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.4 into vk-gl-cts/vulkan-cts-1.1.5
Change-Id: Idea5f48ee1eb6dfa2290d738cabff96c7e88daad
Alexander Galazin [Mon, 18 Nov 2019 10:06:07 +0000 (11:06 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.3 into vk-gl-cts/vulkan-cts-1.1.4
Change-Id: I9adaa87d0bf5d06a12bfc2fcd4b8a30568cefb48
Alexander Galazin [Thu, 14 Nov 2019 16:28:11 +0000 (17:28 +0100)]
Whitelist Vulkan CTS 1.1.6.2
Affects: dEQP-VK.api.driver_properties.properties
Components: Vulkan
Change-Id: I88e01d7a9c754c8a11cb5f4505257fd338eabbf7
Lionel Landwerlin [Wed, 13 Nov 2019 22:05:31 +0000 (00:05 +0200)]
Do not reset perf queries in multi submitted cmdbuffer
Some implementations can require multiple passes to gather values for
their performance queries. It's important not to reset the query pool
in a batch that can be submitted multiple times, otherwise the queries
result risk to never become available.
Compontents: Vulkan
Affects: dEQP-VK.query_pool.performance_query.*
Change-Id: Ief00457fc0808a7c5ee7c0bf6aaf7de7f3042ca5
Lionel Landwerlin [Tue, 15 Oct 2019 15:22:38 +0000 (18:22 +0300)]
Update VK_KHR_performance_query enumerate
Components: Vulkan
Affects: dEQP-VK.query_pool.performance_query.enumerate*
Change-Id: I443c1c9b7a1903063b27b77fc47cbb88259aa407
(cherry picked from commit
bc519fecec3d1b27e0e5ee1a4947257260202eff)
Boris Zanin [Thu, 16 Aug 2018 08:55:01 +0000 (10:55 +0200)]
Tests for VK_KHR_performance_query
Add tests:
* dEQP-VK.query_pool.performance_query.*
Components: Vulkan
VK-GL-CTS issue: 1302
Update VK_KHR_performance_query headers
Align CTS header for VK_KHR_performance_query with API MR 3268
Components: Vulkan
Affects:
dEQP-VK.query_pool.performance_query.*
dEQP-VK.api.info.get_physical_device_properties2.features
dEQP-VK.api.info.get_physical_device_properties2.properties
VK-GL-CTS issue: 1947
Change-Id: I4d3c62dace2d259061f78d3014deb05bc4eac9d7
(cherry picked from commit
32556493fcafd676ec5d8fcfd3492a8c30fa0f97)
(cherry picked from commit
57bdbad1c7d6d527a30ae211164edab945818def)
Ricardo Garcia [Tue, 12 Nov 2019 12:22:22 +0000 (13:22 +0100)]
Check renderpass2 support in image clearing tests
This commit makes sure VK_KHR_create_renderpass2 is supported before
using it in some image clearing tests.
Affected tests:
dEQP-VK.api.image_clearing.*separate_layouts*
Components: Vulkan
VK-GL-CTS issue: 2101
Change-Id: Ib1a44b44f771bb31fd1ef798d2727370d143a1fa
Lionel Landwerlin [Wed, 13 Nov 2019 16:04:22 +0000 (18:04 +0200)]
Fix check of api features
Fixes the cherry pick
282341d2cf5da0d9aaea1547958eb54c3f84ee0f.
Compontents: Vulkan
Affects: dEQP-VK.api.info.get_physical_device_properties2.features
Change-Id: I1d102716767c7bef0b872a5ebe1eb1d77bcd11b5
Alexander Galazin [Wed, 13 Nov 2019 15:17:00 +0000 (16:17 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I96bc5705c41b8ec87b7f14afc0cfa354b84cad4e
Mike Byrne [Fri, 1 Nov 2019 16:50:17 +0000 (16:50 +0000)]
Split descriptor set indexing render pass workload
The render pass workload for descriptor set indexing tests which call
DynamicBuffersInstance::iterate should be split into multiple
submissions to prevent timeouts observed in
dEQP-VK.descriptor_indexing.uniform_buffer_dynamic_in_loop
test.
A number (16) of command buffers, each for a separate render area, are
created and submitted separately.
Affects: dEQP-VK.descriptor_indexing*
Components: Vulkan
VK-GL-CTS Issue: 2061
Change-Id: Ib232854a1224aadbd63eaad30e15f893eeb48f85
Investigation for Khronos issue;
https://gitlab.khronos.org/Tracker/vk-gl-cts/issues/2061
Change-Id: Ib232854a1224aadbd63eaad30e15f893eeb48f85
Tom Cooper [Mon, 28 Oct 2019 15:04:36 +0000 (15:04 +0000)]
Reduce dEQP-VK.pipeline.render_to_image allocs to within platform limits
These tests allocate an image of upto 25% of VkMemoryHeap heapSize and
should be limited to the value in
PlatformMemoryLimits.totalSystemMemory.
Component: Vulkan
VK-GL-CTS Issue: 2087
Affects:
dEQP-VK.pipeline.render_to_image*
Change-Id: Ie7d775cc55fdb54e448fc1c973d1e17925711e47
Tom Cooper [Sat, 26 Oct 2019 18:14:42 +0000 (19:14 +0100)]
Fix descriptor_indexing.input_attachment subpass description
The input attachments in the subpass description did not match the
descriptor set layout and shader. Unused attachment references have
been inserted to pad out the subpass description.
Components: Vulkan
VK-GL-CTS Issue: 2072
Affects: dEQP-VK.descriptor_indexing.input_attachment*
Change-Id: Iab36be779007828ddb25fe0600305a19be448804
Alexander Galazin [Fri, 1 Nov 2019 14:05:53 +0000 (15:05 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.4 into vk-gl-cts/vulkan-cts-1.1.5
Change-Id: I74696bcb3e23aa3e8bde59ebea3137b262dd1394
Alexander Galazin [Fri, 1 Nov 2019 11:16:47 +0000 (12:16 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.3 into vk-gl-cts/vulkan-cts-1.1.4
Change-Id: Ia5d78f0656fb1a2b2e4f989ba556edc3154856f7
Piers Daniell [Tue, 30 Apr 2019 22:50:00 +0000 (16:50 -0600)]
Tests for VK_KHR_separate_depth_stencil_layouts
Test coverage for VK_KHR_separate_depth_stencil_layouts. See
https://gitlab.khronos.org/vulkan/vulkan/merge_requests/3172
This change modifies several test groups to use a single
image aspect bit for depth+stencil formats rather than the
VU required D+S aspects.
Where appropriate the tests are also modified to use
the new depth-only or stencil-only optimal layouts.
Affects:
dEQP-VK.api.copy_and_blit.*d16_unorm_s8_uint*
dEQP-VK.api.copy_and_blit.*d24_unorm_s8_uint*
dEQP-VK.api.copy_and_blit.*d32_sfloat_s8_uint*
dEQP-VK.api.image_clearing*d16_unorm_s8_uint*
dEQP-VK.api.image_clearing*d24_unorm_s8_uint*
dEQP-VK.api.image_clearing*d32_sfloat_s8_uint*
dEQP-VK.pipeline.stencil*d16_unorm_s8_uint*
dEQP-VK.pipeline.stencil*d24_unorm_s8_uint*
dEQP-VK.pipeline.stencil*d32_sfloat_s8_uint*
dEQP-VK.pipeline.depth*d16_unorm_s8_uint*
dEQP-VK.pipeline.depth*d24_unorm_s8_uint*
dEQP-VK.pipeline.depth*d32_sfloat_s8_uint*
dEQP-VK.*.depth_stencil_resolve*d16_unorm_s8_uint*
dEQP-VK.*.depth_stencil_resolve*d24_unorm_s8_uint*
dEQP-VK.*.depth_stencil_resolve*d32_sfloat_s8_uint*
Change-Id: I147f3736614cc897f1319c5aac5ed1fdba2d7b66
Compontents: Vulkan
(cherry picked from commit
39e17004ec33f0e546e6f391f933fb938e8c507d)
Alexander Galazin [Fri, 25 Oct 2019 15:31:07 +0000 (17:31 +0200)]
Whitelist Vulkan CTS 1.1.6.1
Affects: dEQP-VK.api.driver_properties.properties
Components: Vulkan
Change-Id: Id79758cf4688ec414176433edae0914826f088df
Alexander Galazin [Thu, 31 Oct 2019 10:46:59 +0000 (11:46 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I592b6674f5e3a5a1f2a6d050dcc7263b0e5d682c
Dae Kim [Wed, 30 Oct 2019 12:32:17 +0000 (12:32 +0000)]
Add missing write to PointSize
The primitive topology is POINT_LIST for test cases covering the vertex
stage, but the vertex shader in these cases isn't supplying a value for
PontSize.
Affects:
dEQP-VK.binding_model.buffer_device_address.*.vert
Components: Vulkan
VK-GL-CTS issue: 2080
Change-Id: Ibbbeb143bf957f9494b1df84e69bc4a01c891c6a
John Corbally [Wed, 30 Oct 2019 11:14:45 +0000 (11:14 +0000)]
Fix missing image memory barrier in WSI display timing tests
The key part of the missing image memory barrier is changing
VK_IMAGE_LAYOUT_UNDEFINED to VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL
(for the first time).
Affects: dEQP-VK.wsi.android.display_timing.*
Components: Vulkan
VK-GL-CTS issue: 2079
Change-Id: I79e142c9c2a602e066a51e69e3aca9a1d2c0e650
Alexander Galazin [Tue, 29 Oct 2019 13:13:41 +0000 (14:13 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I34ada51e1b9c00280fb9f9d60463989e05f11953
Alexander Galazin [Mon, 28 Oct 2019 15:18:03 +0000 (16:18 +0100)]
Add timelineSempahore to mandatory ext features
https://gitlab.khronos.org/vulkan/vulkan/merge_requests/3421
clarifies that timelineSemaphore is mandatory for
implementations exposing VK_KHR_timeline_semaphore
This CL updates the test to align with the MR.
Components: Vulkan
Affects: dEQP-VK.api.info.device.mandatory_features
Change-Id: I12f9c9e6b05a77109cb0207b748cc9ccfa872d9b
(cherry picked from commit
33cc61974078e64d66e64ccf67ecc906f7a7f3fb)
Alexander Galazin [Mon, 28 Oct 2019 18:51:20 +0000 (19:51 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I3fd2e50726e016645834052ba95fa761f84c5686
Piotr Byszewski [Mon, 28 Oct 2019 15:09:36 +0000 (16:09 +0100)]
Fix check in testMandatoryExtensions
Affects:
dEQP-VK.api.info.android.mandatory_extensions
Change-Id: Iec3b405cb168daa7dd7fc2183963778ef7d6eb0b
Components: Vulkan
Pawel Ksiezopolski [Mon, 28 Oct 2019 14:03:51 +0000 (15:03 +0100)]
Remove deprecated std::bind2nd() call
std::bind2nd() was deprecated in C++11 and removed in C++17.
Affects:
dEQP-VK.descriptor_indexing.*
Components: Vulkan
VK-GL-CTS issue: 2074
Change-Id: I408a39ce9f1632515e3ff6bcfb95966e19aa4b22
Alexander Galazin [Sun, 27 Oct 2019 18:28:49 +0000 (19:28 +0100)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I21178b89019aad73d0b1862e10d5590361fcdcc8
Slawomir Cygan [Thu, 24 Oct 2019 15:45:28 +0000 (17:45 +0200)]
Add scissor matching renderArea to descriptor indexing tests
This fixes rendering outside of render area, what
is forbidden by spec.
This also fixes regression from
7b7ba4e0, where
looping the RenderPass caused workload execution time
to increase and trigger OS-specific device timeout.
Components: Vulkan
VK-GL-CTS Issue: 2069
Affects: dEQP-VK.descriptor_indexing.*
Change-Id: I40f665ff892d03004860583029eb393e09c1ae95
Alexander Galazin [Fri, 25 Oct 2019 16:07:46 +0000 (18:07 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: I365220265de5dd5b239cca76b1ffcd0221404824
Alexander Galazin [Fri, 25 Oct 2019 14:29:30 +0000 (16:29 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.4 into vk-gl-cts/vulkan-cts-1.1.5
Change-Id: Id21fae15bb0fad00a22e2d5798f958e6640804e2
Alexander Galazin [Fri, 25 Oct 2019 11:58:22 +0000 (13:58 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.3 into vk-gl-cts/vulkan-cts-1.1.4
Change-Id: I58c329c990cacdb992c4c9be0b95cb65aa469a80
Piotr Byszewski [Tue, 15 Oct 2019 09:06:51 +0000 (11:06 +0200)]
Check available functionality with context method
This change adds Context::isDeviceFunctionalitySupported and
Context::isInstanceFunctionalitySupported methods that replace
isDeviceExtensionSupported/isInstanceExtensionSupported.
Components: Framework
VK-GL-CTS issue: 2005
Affects:
dEQP-VK.*
Change-Id: If796fc31fe6e769495cdf1ccefa4ae8192caf2b5
Ricardo Garcia [Wed, 23 Oct 2019 16:11:52 +0000 (18:11 +0200)]
Remove NaNs from double compare tests
Affected tests:
dEQP-VK.spirv_assembly.instruction.*.64bit_compare.double.*
Components: Vulkan
VK-GL-CTS issue: 2066
Change-Id: I8d735bbffe2db5d2c8cfac5a616efd6010057408
Arkadiusz Sarwa [Wed, 23 Oct 2019 15:15:12 +0000 (17:15 +0200)]
Fix incorrect assumption for pipeline cache hit
The test incorrectly assumes that adding a pipeline creation flag bit
VK_PIPELINE_CREATE_ALLOW_DERIVATIVES_BIT to only one pipeline
still should cause a cache hit. Fixed the problem by adding the
VK_PIPELINE_CREATE_ALLOW_DERIVATIVES_BIT flag bit to both pipelines.
Affects:
dEQP-VK.pipeline.creation_feedback.compute_tests.compute_stage
Components: Vulkan
VK-GL-CTS issue: 2068
Change-Id: Idb963219a4e894eee6a0bb3b1fc20797d5b0323e
Alexander Galazin [Tue, 22 Oct 2019 16:59:45 +0000 (18:59 +0200)]
Merge vk-gl-cts/vulkan-cts-1.1.5 into vk-gl-cts/vulkan-cts-1.1.6
Change-Id: Ic9de3ec00dbd38b2106f1c497059e40d4d361297
Lionel Landwerlin [Wed, 16 Oct 2019 18:39:50 +0000 (21:39 +0300)]
Fix waitSemaphoreCount value in signal ordering tests
VK-GL-CTS Issue: 2059
Components: Vulkan
Affects: dEQP-VK.synchronization.signal_order.*timeline_semaphore.*
Change-Id: I25b86f1ae2b8e3650e061f9f2e345147c01419c6