Vulkan CTS 1.1.3.0
Vulkan CTS 1.1.3.0
New tests:
- Tests for VK_KHR_depth_stencil_resolve
- Tests for VK_EXT_conditional_rendering
- Tests for VK_EXT_vertex_attribute_divisor
- Tests for different data types for workgroup memory
- Tests to intentionally mismatch opImageType
- Tests using VkRunner, precompiling the scripts
- Add subpass external dependency to pushDescriptor tests
- Add tests for varying naming
- Added tests for OpCompositeInsert
- Enable remaining OpFUnord tests
- Add test veryifying if proper 'count' is written in wsi getters
- Add barrier at start of each CB to order wrt previous CBs
- Enable previously excluded acos test
- Add tests for derivatives in dynamic but quad-uniform control flow
- Add tests for booleans produced from different bit-sizes
- Add tests for late fragment test dependencies
- Extend push desriptors test to 128 compute dispatches
- Extend transcoding tests to 1D/3D images
- Test 2D texture swizzles with VkComponentMappings
- Test depth/stencil resolve attachments
- Added tests for out-of-order attribute locations
- test layered resolves with a single region convering all layers.
- Test storing protected data in workgroup memory
- Added tests for non-linear attribute locations
- Add corner case values for float32 to int16 conversion tests.
- Add tests for subpass dependencies
- Test clip/cull distance reading in fragment shader
- New tests for descriptor dynamic offset
- Add dependency checking to dEQP-VK.api.info.*.extensions
- Added several abuses of OpName and OpMemberName
- dEQP-VK: Add tests for blitting to/from linear images
- Test OpPtrAccessChain with base alignment of < 16
- Test render pass attachments not referenced by subpasses
- Test cube image blits with VK_IMAGE_CREATE_CUBE_COMPATIBLE flag.
- Simultaneous depth+stencil components blit test
- Ensure there's no KHX extensions
- Added test set for unnormalized coord samplers
- Add secondary command buffers to pipeline barrier tests
- Create Compute test for OpCompositeExtract
- New test for copying 16bit input to two outputs.
- Test push descriptors
Fixes:
- Fixes multiview query tests
- Fix dependency on unspecified rounding mode
- Correct support conditions for float16s
- Clear Core Validation Errors for Protected Memory
- Fix dup tests to not free in-use semaphores/fences
- Fix 0b prefix in conditional rendering tests
- Fix image aspects in multiview tests
- Fix memory requirements for draw_clear tests
- Fix further aspect masks in renderpass2 tests
- Simplify YCbCr result verification
- Fix initial YCbCr image layout
- Fix validation of occlusion query tests
- Fix more memory flush/invalidation cases
- Validation fixes for device_init tests
- Be more careful about Block decorations on structs
- Remove duplicate tests under dEQP-VK.renderpass*
- Fix API usage errors in fragment operation tests
- Fixed incorrect load/store ops for depth in multiview test
- Reduce devices created in device_group tests
- Remove duplicate identical make*MemoryBarrier
- Allow SRGB conversion for store tests to generate reference image.
- Fix subgroup shuffle tests to do testing
- Remove duplication from subgroup shuffle tests
- Set the correct "unused" attachment initialLayout
- Validation: Use correct aspects in renderpass2
- Fix leak in shaderCacheLoad()
- Correct image mutable tests on null driver
- Remove duplicate float16Int8Features
- Remove StorageUniform16 from SPIR-V conversion tests
- Fix anonymous namespace in UBO matrix padding test
- Properly align and pad vec16x2 values in UBO array
- Require VK_KHR_shader_float16_int8 in float16 opconstant tests
- Fixes for Protected Memory
- Further validation fixes for renderpass tests
- Fix invalid color values in image.mutable tests
- Use the correct type for image creation flags
- Fixes AHB image tests feature check
- Added framework function for image to buffer copy
- Correct invalidate->flush after writing to host-mapped memory
- Specify PointSize when drawing point lists
- Utilize copyBufferToImage framework function where possible
- Fixed unused attachment tests to properly use renderpass2
- SubgroupBallot: Don't test 0 <= id
- Rationalise the 'check*' functions in subgroups
- Remove one copy of the clustered test sources
- Further validation fixes for renderpass tests
- Use GENERAL image layout for images used as STORAGE image
- Properly align and pad vec16x2 values in UBO array
- Fix for unnormalized coordinate test using mipmaps
- Fix API usage errors in dEQP-VK.geometry.basic.*
- Fix swapchain creation in dEQP-VK.wsi..swapchain.
- Require VK_KHR_shader_float16_int8 in float16 opconstant tests
- Import dedicated allocs w/ importDedicatedMemory
- Properly analyze edge0>=edge1 cases in smoothstep
- Remove NaN from OpFOrd* and OpFUnord* tests
- Validation fixes for device group tests
- Fix validation in dynamic offset tests
- Mark storage buffer as readonly
- Fix validation in renderpass tests
- Fix precision mismatches in GLSL tests
- Place OpVariable at the beginning of the block
- Fix barrier range in UpdateBuffer tests
- Fix validation in memory tests
- Fix memory flush alignments in binding model tests
- Fix validation in command buffer tests
- Fix validation in memory binding tests
- Fix queue validation in synchronization tests
- Fix non-valid barriers in dEQP-VK.tessellation.*
- Check sparse support in memory requirements tests
- Don't pass tess and geom flags when not needed
- Fix testing of various device limits
- Fix api.smoke.* tests for validation
- Fix unused function warnings in the CTS code
- Add missing invalidates in renderpass tests
- Fix synchronisation in ycbcr tests
- Use a valid final layout in renderpass tests
- Fix valid usage for SampleLocations extension
- Fix some pipeline and memory barriers
- Fix DeviceFeatures infinite loop in pNext chain
- Fix array stride in uniform buffer for VK_KHR_16bit_storage
- Drop superfluous Clip/CullDistance capabilities
- Remove invalid execution mode in smoke test
- Fix array stride in uniform buffer for VK_KHR_16bit_storage
- Fix asserts in image.mutable tests
- Remove over-zealous sanity check for memory types
- dEQP-VK: Stop requiring the destination format to support linear sampling for blit tests
- Fix copy/paste errors in opundef assembly tests
- Fixes geometry shader invocation dependency
- Fix malformed assets, use DE_FATAL
- Add Flat decorations to integer fragment shader inputs
- Fix result checking in multisample resolve tests
- Fix asserts and check for base image format support in image.mutable tests
- Fix NaN handling in funord tests
- Fix validation failure in float_controls
- Fixes ycbcr image properties requirement
- Fix image type in Uniform texel buffer tests
- Protected Memory Fixes - Barrier Layouts & Flags
- Update VK_KHR_vulkan_memory_model to spec version 3
- Fixes depth load_op in multiview tests
- Truncate random coordinates in late fragment tests
- Touch watchdog more often for precision tests.
- Added PointSize in shader used for drawing points
- Add AHB to Android allowed extensions list
- Fix logging of 8-bit integer values
- Fix incorrect use of copyImageToBuffer in TextureRenderer::renderQuad()
- Allow coarse derivatives to be calced by fine func
- Added missing writes to gl_Position in vertex shaders
- Depth attachment not stored in late_fragment_tests
Improvements:
- Compile shaders for 1.0-compatible tests as SPIRV 1.0, not max supported
- Handle gcc 7 warnings
- Build prebuilts for HLSL shaders
- Remove one method of specifying features in vktSpvAsm
- Cause vkrunner tests to fail if scripts not found
- Allow a tolerance for results of sqrt
- Add framework function createInstanceWithExtensions
- Add support for RenderDoc in Vulkan tests
- Added script for vulkan.h.in generation
- Use precompiled headers on MSVC
- Remove test lib dependencies on deqp-vk-package.
- Add a flushAlloc helper to fix valid usage
- Add SPIR-V optimization options to vk-build-programs
- Improve wsi.incremental_present.* for low memory