iostrows [Thu, 7 Jan 2016 15:06:15 +0000 (16:06 +0100)]
Do not pass signaled fence to vkQueueSubmit()
Pyry Haulos [Wed, 6 Jan 2016 21:51:54 +0000 (13:51 -0800)]
Merge in changes from Khronos repository
Pyry Haulos [Wed, 6 Jan 2016 19:33:44 +0000 (14:33 -0500)]
Merge branch 'mem-random-div-zero' into 'master'
Fix division by zero in random memory allocation test
I got a division by 0 with some configuration when running some random memory allocation tests.
The related code is:
const VkDeviceSize size = 1 + (rng.getUint64() % ((m_heap.size / MAX_MEMORY_USAGE_DIV) - m_usage - 1));
It's possible for "(m_heap.size / MAX_MEMORY_USAGE_DIV) - m_usage - 1" to be equal to zero since we are doing a -1 and the method "full" just compare "m_usage * MAX_MEMORY_USAGE_DIV >= m_heap.size".
With this change I do a max(1) which should resolve the issue
See merge request !228
Pyry Haulos [Wed, 6 Jan 2016 19:31:49 +0000 (11:31 -0800)]
Change maxAnisotropy to 1 in object mgmt and smoke tests
Anisotropic filtering is optional and value 1 disables it.
Change-Id: Ib1d5ff3e72d8271eb24321c04782dbe9cd9c833f
Pyry Haulos [Wed, 6 Jan 2016 19:16:32 +0000 (14:16 -0500)]
Merge branch 'jbolz_copy_and_blit_fixes' into 'master'
Copy_and_blit fixes: replace format during CopyImage expected result calculation…
…, to make it use memcpy. Use Any memory for image allocation for buffer-to-image copy.
See merge request !234
Pyry Haulos [Wed, 6 Jan 2016 19:13:09 +0000 (11:13 -0800)]
Merge changes from Khronos repository
Pyry Haulos [Wed, 6 Jan 2016 18:57:20 +0000 (13:57 -0500)]
Merge branch 'jbolz_shaderrender_128' into 'master'
Change 'shaderrender' (glsl) tests to use a size of 128x128 rather than 100x100,…
… to avoid interpolation noise.
This is a slightly unusual bugfix, but I was seeing failures in the dEQP-VK.glsl.operator.binary_operator.bitwise_xor and dEQP-VK.glsl.operator.float_compare.lessThan/lessThanEqual (and GT/GEQ) tests, only in the fragment shader variants. These are both tests where a small interpolation difference can get magnified into a very large diff due to the nature of the operation, and this was causing these tests to fail. Using power-of-two dimensions reduces the frequency of GPU vs reference interpolation differences.
See merge request !225
Pyry Haulos [Wed, 6 Jan 2016 17:51:56 +0000 (12:51 -0500)]
Merge branch 'awoloszyn-fix-undefined-ids' into 'master'
Fixes a handful of tests that used undefined IDs.
Fixes #137
Fixes #149
See merge request !243
Pyry Haulos [Wed, 6 Jan 2016 17:51:40 +0000 (12:51 -0500)]
Merge branch 'jbolz_bitwisexor' into 'master'
Fix BitwiseXor specconstant test. Use only positive values for SRem/SMod
See bugs https://cvs.khronos.org/bugzilla/show_bug.cgi?id=15380 and https://cvs.khronos.org/bugzilla/show_bug.cgi?id=15379. Note that we will need to pick up an updated SPIRV-tools with the fix for 15379.
See merge request !235
Pyry Haulos [Wed, 6 Jan 2016 17:51:07 +0000 (12:51 -0500)]
Merge branch '116-bufferview-swizzle-scale-and-bias' into 'master'
Enhance verification of swizzling in pipeline.image_view tests
Fixes issue #116
- apply swizzle also to scale and bias parameters in shaders, so valid channels are not zeroed
Test pipeline.image_view verify the functionality of sampling a view of image by performing following operation:
```
texture(texSampler, vtxTexCoords) * lookupScale + lookupBias
```
and comparing results from Vulkan and ref renderer. The problem here is lookupScale and lookupBias ignore the swizzle set in image view object, thus for non 4-channel formats scale and bias zero the result on wrong channel, causing loss on information and comparing 0-es to 0-es in final validation of rendered image.
As a fix, I propose to apply component mapping from image view to both values (both in Vulkan and ref renderer shader).
See merge request !192
Pyry Haulos [Wed, 6 Jan 2016 17:50:13 +0000 (12:50 -0500)]
Merge branch '98-remove-3d-astc' into 'master'
Removed test cases for 3D ASTC textures.
See merge request !242
Pyry Haulos [Wed, 6 Jan 2016 17:47:29 +0000 (12:47 -0500)]
Merge branch 'jbolz_pipeline_barrier_fixes2' into 'master'
pipeline_barrier: Keep the VkBuffer in the PrepareContext alive until the work is executed...
…, by putting it in the same scope as the execution. Also wait for idle before the destructors are called.
This may not be the best fix, but it works. Buffers can be created and destroyed by various OPs in the test, but I think all the Execute work happens with the last buffer. The test may need some restructuring to using the different buffers.
See merge request !230
Andrew Woloszyn [Wed, 6 Jan 2016 16:33:28 +0000 (11:33 -0500)]
Fixes a handful of tests that used undefined IDs.
Fixes #137
Fixes #149
Dae Kim [Wed, 6 Jan 2016 10:47:08 +0000 (10:47 +0000)]
Removed test cases for 3D ASTC textures.
Dan Willemsen [Wed, 6 Jan 2016 00:40:49 +0000 (00:40 +0000)]
Merge "Use -Wno-date-time to allow __DATE__/__TIME__" am:
bd1826b4a3 am:
f4e3b26175
am:
c3d988aab7
* commit '
c3d988aab70b74c98ce51d7345eac519ad5c9222':
Use -Wno-date-time to allow __DATE__/__TIME__
Dan Willemsen [Wed, 6 Jan 2016 00:38:56 +0000 (00:38 +0000)]
Merge "Use -Wno-date-time to allow __DATE__/__TIME__" am:
bd1826b4a3
am:
f4e3b26175
* commit '
f4e3b26175490978db76e76576b488cbe2cf6bb6':
Use -Wno-date-time to allow __DATE__/__TIME__
Dan Willemsen [Wed, 6 Jan 2016 00:36:54 +0000 (00:36 +0000)]
Merge "Use -Wno-date-time to allow __DATE__/__TIME__"
am:
bd1826b4a3
* commit '
bd1826b4a347e28e7aab1479d53b1c6d8ce4c640':
Use -Wno-date-time to allow __DATE__/__TIME__
Dan Willemsen [Wed, 6 Jan 2016 00:23:32 +0000 (00:23 +0000)]
Merge "Use -Wno-date-time to allow __DATE__/__TIME__"
Pyry Haulos [Tue, 5 Jan 2016 21:00:04 +0000 (13:00 -0800)]
Merge changes from Khronos repository
Pyry Haulos [Tue, 5 Jan 2016 20:54:26 +0000 (12:54 -0800)]
Merge 138-dont-require-tess-geom
Pyry Haulos [Tue, 5 Jan 2016 20:34:42 +0000 (20:34 +0000)]
Merge "Update MNC GLES2 mustpass list"
am:
661fef704e
* commit '
661fef704eeecc3d83997bffc2615b19e68efa81':
Update MNC GLES2 mustpass list
Pyry Haulos [Tue, 5 Jan 2016 20:33:22 +0000 (20:33 +0000)]
Merge "Update MNC GLES2 mustpass list"
David Sodman [Tue, 5 Jan 2016 20:06:16 +0000 (20:06 +0000)]
Merge "Add check for unsupported limits in api.info.device" into vulkan
Pyry Haulos [Tue, 5 Jan 2016 20:00:35 +0000 (20:00 +0000)]
Remove few tests affected by a HW limitation from master and MNC lists
am:
2c2db488c8
* commit '
2c2db488c8596e3c8fa9791d019aee22cd654821':
Remove few tests affected by a HW limitation from master and MNC lists
Pyry Haulos [Tue, 5 Jan 2016 19:47:58 +0000 (11:47 -0800)]
Update MNC GLES2 mustpass list
Change-Id: I9a123692804d61fd716e39b60a87276baeb3fcc9
Pyry Haulos [Tue, 5 Jan 2016 19:46:37 +0000 (11:46 -0800)]
Remove few tests affected by a HW limitation from master and MNC lists
Some primitive_bbox tests were earlier removed due to a HW defect, but
the list was incomplete.
Bug:
22713865
Change-Id: I989e5c65de75f2b67b3e19f17a531b37831f0fe9
Pyry Haulos [Tue, 5 Jan 2016 19:12:36 +0000 (11:12 -0800)]
Clean up test hierarchy and case naming
Affects following test groups:
* dEQP-VK.glsl.ShaderExecutor
* dEQP-VK.spirv_assembly.instruction.graphics
Change-Id: Ia14ee2deb112670d0a222380b5e0ddf6e03daffd
Pyry Haulos [Tue, 5 Jan 2016 18:57:31 +0000 (13:57 -0500)]
Merge branch 'jbolz_pipeline_barrier_fixes' into 'master'
pipeline_barrier fixes
- Try to use the full (power of two) buffer size. Not doing so tickles several issues to do with not writing all bytes of the buffer or image.
- Fix FillBuffer::verify byte selection.
- Use R8G8 rather than R8G8B8 for the vertex format. The third component is out of bounds for the last vertex, which makes the entire attribute undefined (i.e. substituted with zero on some implementations).
- Subtract out epsilon in vertex shader math to avoid points on the frustum boundary which may be culled.
See merge request !227
Andrew Woloszyn [Tue, 5 Jan 2016 16:52:02 +0000 (11:52 -0500)]
Use context.getDeviceFeatures instead of instanceInterface.
Pyry Haulos [Tue, 5 Jan 2016 16:41:18 +0000 (11:41 -0500)]
Merge branch '144-145-fix-assembly' into 'master'
Fixes for assembly in opMemoryAccess and opQuantize tests.
Fixes #144
Fixes #145
See merge request !241
Pyry Haulos [Tue, 5 Jan 2016 16:40:46 +0000 (11:40 -0500)]
Merge branch '141-fix-android-build-win-64' into 'master'
Fix Android build script on Windows 64 bit (NDK r10e)
- Correct Host OS name on Windows 64 bit in common.py
- Add the missing Host OS in toolchain file for NDK r10e
- Downgrade to LLVM 3.5 in NDK r10e because the prebuilt version 3.6 is
missing the llvm-ar utility which is needed by the build script.
See merge request !236
Andrew Woloszyn [Tue, 5 Jan 2016 15:17:39 +0000 (10:17 -0500)]
Updated to only include necessary stages in tests.
This simplifies the framework, and flow of the tests.
Any stage present in a test will be required. Tests will
only require tess/geom if we are specifically testing those stages.
Andrew Woloszyn [Tue, 5 Jan 2016 14:52:28 +0000 (09:52 -0500)]
Fixes for assembly in opMemoryAccess and opQuantize tests.
Fixes #144
Fixes #145
Maciej Jesionowski [Mon, 4 Jan 2016 09:45:10 +0000 (10:45 +0100)]
Fix Android build script on Windows 64 bit (NDK r10e)
- Correct Host OS name on Windows 64 bit in common.py
- Add the missing Host OS in toolchain file for NDK r10e
- Downgrade to LLVM 3.5 when using Windows NDK r10e,
because the prebuilt version 3.6 is missing the llvm-ar
utility which is needed by the build script.
David Sodman [Mon, 28 Dec 2015 21:33:39 +0000 (13:33 -0800)]
Add check for unsupported limits in api.info.device
Fixes https://gitlab.khronos.org/vulkan/vulkancts/issues/120
Change-Id: If43e22c7007325eb00d87bbe59a2dca65aae584e
Jeff Bolz [Sun, 3 Jan 2016 21:20:52 +0000 (21:20 +0000)]
Fix BitwiseXor specconstant test. Use only positive values for SRem/SMod
Pyry Haulos [Mon, 4 Jan 2016 22:00:05 +0000 (17:00 -0500)]
Merge branch 'bug137-fix-spirv_assembly.instruction.graphics.module-geom-var-decls' into 'master'
Fix 137: bad geom entry point I/O var declarations
Fix declarations of Input and Output declarations in test
dEQP-VK.spirv_assembly.instruction.graphics.module.*
They were missing "geom_" prefices.
Changed:
OpEntryPoint Geometry %geom_main "main" %out_gl_position %gl_in
%out_color %in_color
To the following, to match usage in the function bodies:
OpEntryPoint Geometry %geom_main "main" %geom_out_gl_position
%geom_gl_in %geom_out_color %geom_in_color
Also fix associated OpDecorate instructions.
Fixes #137
See merge request !238
Pyry Haulos [Mon, 4 Jan 2016 21:54:03 +0000 (13:54 -0800)]
Fix implicit VkDeviceSize -> size_t conversion warnings
Change-Id: Ibb4b63cf0651ddec66a62e39cc1ab1b155cedd13
Pyry Haulos [Mon, 4 Jan 2016 21:25:50 +0000 (13:25 -0800)]
Fix a few warnings
Change-Id: Icc60dcf8f72c1b9140c76bb1568af37b0b3b1843
Pyry Haulos [Mon, 4 Jan 2016 20:56:21 +0000 (12:56 -0800)]
Merge in changes from Khronos repository
Andrew Woloszyn [Wed, 30 Dec 2015 21:57:09 +0000 (16:57 -0500)]
spirv_assembly.instruction.graphics.* tests no long require tess/geom.
Tessellation/Geoemtry shaders will be used if present, but disabled
otherwise. NotSupported is set for any tests that requires an optional
feature.
Pyry Haulos [Mon, 4 Jan 2016 19:16:02 +0000 (14:16 -0500)]
Merge branch 'shaderExecutor-imageLayout' into 'master'
ShaderExecutor: fixed FragmentOutExecutor that used Images in UNDEFINED layout
Added proper transitions and fixed layout used for cmdCopyImageToBuffer
See merge request !224
Pyry Haulos [Mon, 4 Jan 2016 19:14:37 +0000 (14:14 -0500)]
Merge branch 'scygan_image_reinterpret_snorm_magic_0x80' into 'master'
dEQP-VK.image.reinterpret: bad handling of -128 SNORM value
Some tests in reintepret group create images of R32I type, than issue an imageLoad()-ed in shader, reinterpreting the image as RGBA8_SNORM. Than the value is imageStore()-ed again as RGBA8_SNORM.
Effectively each channel undergoes SNORM8 -> float -> SNORM8 conversion.
The problem is test requires the stored image to match exactly the loaded one, while the loaded image contains bytes of -128 value.
-128 (same as 127) converts to -1.0 float, however according to following equation in spec, the -1.0 float value can convert only to 0x81, what makes test fail.
```
f′=convertFloatToInt(f*(2^(b−1)−1),b)
```
The proposed fix here is to replace all 0x80 values in loaded image - same it is already done for NaNs and Infs for float format.
See merge request !233
David Neto [Mon, 4 Jan 2016 15:33:32 +0000 (10:33 -0500)]
Fix 137: bad geom entry point I/O var declarations
Fix declarations of Input and Output variable declarations and
the gl_PerVertex type declaration in test
dEQP-VK.spirv_assembly.instruction.graphics.module.*
They were missing "geom_" prefices.
Changed:
OpEntryPoint Geometry %geom_main "main" %out_gl_position %gl_in
%out_color %in_color
To the following, to match usage in the function bodies:
OpEntryPoint Geometry %geom_main "main" %geom_out_gl_position
%geom_gl_in %geom_out_color %geom_in_color
Also fix associated OpDecorate instructions.
Should fix https://gitlab.khronos.org/vulkan/vulkancts/issues/137
scygan [Mon, 28 Dec 2015 19:11:56 +0000 (20:11 +0100)]
ShaderExecutor: fixed FragmentOutExecutor that used Images in UNDEFINED layout
Added proper transitions and fixed layout used for cmdCopyImageToBuffer
mchomicz [Wed, 30 Dec 2015 13:08:19 +0000 (14:08 +0100)]
replace problematic values 0x80 (-128) for snorm formats
Pyry Haulos [Sun, 3 Jan 2016 22:03:43 +0000 (17:03 -0500)]
Merge branch '134-fix-spirv-graphics-loop' into 'master'
Fixed invalid spirv in dEQP-VK.spirv_assembly.instruction.graphics.loop.*
Fix for #134
See merge request !232
Pyry Haulos [Sun, 3 Jan 2016 22:03:23 +0000 (17:03 -0500)]
Merge branch '136-fix-opMemoryAccess' into 'master'
Fixed invalid SPIR-V in spirv_assembly.instruction.graphics.opMemoryAccess.*
For issue #136
See merge request !231
Pyry Haulos [Sun, 3 Jan 2016 22:02:38 +0000 (17:02 -0500)]
Merge branch '133-fix-opfrem' into 'master'
Fix invalid SPIR-V in dEQP-VK.spirv_assembly.instruction.graphics.frem.*
See merge request !229
Pyry Haulos [Sun, 3 Jan 2016 22:01:03 +0000 (17:01 -0500)]
Merge branch '132-push-constant-fix' into 'master'
Fix bugs for vertex attribute count in push constant tests.
Fix issue #132 : set wrong vertex attribute count in pipeline
See merge request !226
Pyry Haulos [Sun, 3 Jan 2016 21:58:23 +0000 (16:58 -0500)]
Merge branch 'shaderExecutor-imageView' into 'master'
ShaderExecutor: fixed FragmentOutExecutor that used imageView with mipLevels == 0
VkImageViewCreateInfo.mipLevels should equal to 1 to render anything
See merge request !222
Pyry Haulos [Sun, 3 Jan 2016 21:57:50 +0000 (16:57 -0500)]
Merge branch 'jbolz_ssbo_fix' into 'master'
Fix descriptors for SSBO tests.
This is the same fix as in !187, just for SSBO tests.
See merge request !220
Pyry Haulos [Sun, 3 Jan 2016 21:56:17 +0000 (16:56 -0500)]
Merge branch 'api-copiesAndBlitting-bufferUsage' into 'master'
copies_and_blitting test: fix case 'buffer_to_image' - changed src buffer usage flags
There is a buffer in 'buffer_to_image' test, that is used as a source of a copy command, however it's usage is set to DST.
See merge request !219
Pyry Haulos [Sun, 3 Jan 2016 21:55:54 +0000 (16:55 -0500)]
Merge branch 'jbolz_fix_floatbitstouint' into 'master'
Fix floatbitstouint shaderexecutor test copy/paste bug
See merge request !217
Pyry Haulos [Sun, 3 Jan 2016 21:55:28 +0000 (16:55 -0500)]
Merge branch 'jbolz_draw_instance_index' into 'master'
Change instanceID shader to use InstanceIndex
Change gl_InstanceID to gl_InstanceIndex. Add location qualifier on fragment input. Increase size of rendertarget to reduce likelihood of false success due to fuzzy compare.
Fixes #123.
See merge request !216
Pyry Haulos [Sun, 3 Jan 2016 21:54:36 +0000 (16:54 -0500)]
Merge branch 'dynamic-state-fixes' into 'master'
Various fixes for dynamic state tests.
* Added location qualifier to vertex output variables.
* Added missing initialization of flags in some CreateInfo structures.
* Check if depth bounds test is supported.
* Check if multi-viewport is supported.
* Check if geometry shaders are supported.
* Fixed reference image in lineWidth test.
See merge request !215
Pyry Haulos [Sun, 3 Jan 2016 21:53:44 +0000 (16:53 -0500)]
Merge branch '122-fix-timestamp-tests-typo' into 'master'
Fix timestamp tests typo (issue #122)
Fix a typo and some assertions to make timestamp basic compute tests work.
See merge request !214
Pyry Haulos [Sun, 3 Jan 2016 21:51:46 +0000 (16:51 -0500)]
Merge branch 'jbolz_fix_cmdbuf' into 'master'
Fix command buffer tests.
Fix number of command buffers for secondary_execute_twice test. Fix result buffer memory binding for order_bind_pipeline test. Fix comments for CommandBufferAllocateInfo structs.
See merge request !213
Pyry Haulos [Sun, 3 Jan 2016 21:50:12 +0000 (16:50 -0500)]
Merge branch 'jbolz_push_constants' into 'master'
Push constants fixes
Push Constant fixes: add location=0 on color I/O. Replace 'size 40' test with 'size 48', replacing two push constant blocks with one and using proper std140 alignment.
See merge request !206
Pyry Haulos [Sun, 3 Jan 2016 21:49:39 +0000 (16:49 -0500)]
Merge branch 'jbolz_ubo_align' into 'master'
Fix 'ubo' tests to lay out data satisfying the min ubo alignment
Fixes #114.
I didn't see any way to get the alignment during the init() function, so I just made a copy of it before download.
See merge request !205
Pyry Haulos [Sun, 3 Jan 2016 21:49:10 +0000 (16:49 -0500)]
Merge branch 'jbolz_glsl_matrix2' into 'master'
Fix glsl.matrix.outerproduct tests
Using 'getPtr' selected a template instantiation that downloaded the pointer value rather than the vector.
See merge request !204
Jeff Bolz [Fri, 1 Jan 2016 17:28:19 +0000 (17:28 +0000)]
Copy_and_blit fixes: replace format during CopyImage expected result calculation, to make it use memcpy. Use Any memory for image allocation for buffer-to-image copy.
Andrew Woloszyn [Wed, 30 Dec 2015 22:51:51 +0000 (17:51 -0500)]
Fixed invalid spirv in dEQP-VK.spirv_assembly.instruction.graphics.loop.*
Andrew Woloszyn [Wed, 30 Dec 2015 22:40:35 +0000 (17:40 -0500)]
Fixed invalid SPIR-V in spirv_assembly.instruction.graphics.opMemoryAccess.*
Jeff Bolz [Wed, 30 Dec 2015 22:25:01 +0000 (22:25 +0000)]
Keep the VkBuffer in the PrepareContext alive until the work is executed, by putting it in the same scope as the execution. Also wait for idle before the destructors are called.
Andrew Woloszyn [Wed, 30 Dec 2015 22:23:04 +0000 (17:23 -0500)]
Fix invalid SPIR-V in dEQP-VK.spirv_assembly.instruction.graphics.frem.*
Jeannot Breton [Wed, 30 Dec 2015 18:35:21 +0000 (12:35 -0600)]
Avoid division by zero
Jeff Bolz [Wed, 30 Dec 2015 09:07:50 +0000 (09:07 +0000)]
pipeline_barrier fixes
Fei Yang [Wed, 30 Dec 2015 00:52:18 +0000 (08:52 +0800)]
Fix bugs for vertex attribute count in push constant tests.
Change-Id: Iccacad6b3a787feedbb0b4df7f8689226fa64a7a
Pyry Haulos [Tue, 29 Dec 2015 22:07:28 +0000 (22:07 +0000)]
Remove few tests affected by a HW limitation am:
3f71117a2f am:
58213c37ff am:
8f5b937334 -s ours
am:
e2eb2f509b
* commit '
e2eb2f509bd1a6a041d6e39fcbe4844708e83336':
Remove few tests affected by a HW limitation
Pyry Haulos [Tue, 29 Dec 2015 21:59:53 +0000 (21:59 +0000)]
Remove few tests affected by a HW limitation am:
3f71117a2f am:
58213c37ff
am:
8f5b937334 -s ours
* commit '
8f5b937334b80c4961fe4119984e23279d911c0a':
Remove few tests affected by a HW limitation
Pyry Haulos [Tue, 29 Dec 2015 21:38:06 +0000 (21:38 +0000)]
Remove few tests affected by a HW limitation am:
3f71117a2f
am:
58213c37ff
* commit '
58213c37ff819fce0572fb0e65b3b9fe3e19b398':
Remove few tests affected by a HW limitation
Pyry Haulos [Tue, 29 Dec 2015 21:29:08 +0000 (21:29 +0000)]
Remove few tests affected by a HW limitation
am:
3f71117a2f
* commit '
3f71117a2f57d08ef9a5efe2f211ba515268af60':
Remove few tests affected by a HW limitation
pflisiko [Mon, 28 Dec 2015 14:40:22 +0000 (15:40 +0100)]
fix case 'buffer_to_image' - changed src buffer usage flags
Dae Kim [Tue, 29 Dec 2015 10:41:20 +0000 (10:41 +0000)]
Various fixes for dynamic state tests.
* Added location qualifier to vertex output variables.
* Added missing initialization of flags in some CreateInfo structures.
* Check if depth bounds test is supported.
* Check if multi-viewport is supported.
* Check if geometry shaders are supported.
* Fixed reference image in lineWidth test.
Jeff Bolz [Tue, 29 Dec 2015 06:33:45 +0000 (06:33 +0000)]
Change 'shaderrender' (glsl) tests to use a size of 128x128 rather than 100x100, to avoid interpolation noise.
Kenneth Russell [Mon, 28 Dec 2015 23:45:36 +0000 (23:45 +0000)]
Added test case for conditional assignments to matrix elements.
am:
cbf84e3898
* commit '
cbf84e38987721d18994470b5ecabd0c2f812590':
Added test case for conditional assignments to matrix elements.
Jeff Bolz [Sun, 27 Dec 2015 20:53:46 +0000 (20:53 +0000)]
Change gl_InstanceID to gl_InstanceIndex. Increase width/height to avoid false positives in fuzzy compare.
scygan [Mon, 28 Dec 2015 19:04:07 +0000 (20:04 +0100)]
ShaderExecutor: fixed FragmentOutExecutor that used imageView with mipLevels == 0
VkImageViewCreateInfo.mipLevels should equal to 1 to render anything
Jeff Bolz [Mon, 28 Dec 2015 19:08:28 +0000 (19:08 +0000)]
Fix descriptors for SSBO tests.
Jeff Bolz [Mon, 28 Dec 2015 07:36:55 +0000 (07:36 +0000)]
Fix floatbitstouint shaderexecutor test copy/paste bug
Rocky Shi [Mon, 28 Dec 2015 03:33:05 +0000 (11:33 +0800)]
Fix issues after code review.
Jeff Bolz [Sun, 27 Dec 2015 21:07:49 +0000 (21:07 +0000)]
Merge branch 'master' into jbolz_push_constants
Pyry Haulos [Sun, 27 Dec 2015 03:48:33 +0000 (16:48 +1300)]
Fix warnings in copies and blit tests
Pyry Haulos [Sun, 27 Dec 2015 03:44:57 +0000 (16:44 +1300)]
Merge 70-copies-and-blitting
Pyry Haulos [Sun, 27 Dec 2015 03:33:17 +0000 (22:33 -0500)]
Merge branch 'dkoch_mv_draw_shaders' into 'master'
Move draw shaders to correct location
The draw/ tests are expecting the path vulkan/draw/ instead of just draw/.
See merge request !212
Pyry Haulos [Sun, 27 Dec 2015 03:30:07 +0000 (22:30 -0500)]
Merge branch 'imageLoadStore-too-small-descriptorPool' into 'master'
vktImageLoadStoreTests: fix *__layer test cases - fix too small descriptor pool
In dEQP-VK.image.store.1d_array.r32g32b32a32_sfloat_single_layer test case, there are 8 descriptor sets allocated from descriptor pool that has space only for one.
See merge request !210
Pyry Haulos [Sun, 27 Dec 2015 03:29:17 +0000 (22:29 -0500)]
Merge branch 'renderpass-proper-aspect-for-cmdClearAttachments' into 'master'
Renderpass test: use proper aspect for cmdClearAttachments
dEQP-VK.renderpass.attachment.3.206 tests uses this chunk of code to clear D32 image, so VK_IMAGE_ASPECT_STENCIL_BIT is invalid.
See merge request !209
Pyry Haulos [Sun, 27 Dec 2015 03:28:09 +0000 (22:28 -0500)]
Merge branch '80-ApiObjectManagement-desriptorPool-size' into 'master'
Fix descriptor pool size in api.object_management.multiple_shared_resources.descriptor_set
Fixes issue #80
- there are multiple sets allocated, so desciptor VkDescriptorPoolSize.descriptorCount must be multipled by number
of sets
See merge request !208
Pyry Haulos [Sun, 27 Dec 2015 03:27:28 +0000 (22:27 -0500)]
Merge branch 'api-bufferView-image-layout' into 'master'
vktApiBufferViewAccessTests: image cannot be created with COLOR_ATTACHMENT_OPTIMAL layout.
- According to spec, the only valid layouts in vkCreateImage are UNDEFINED and PREINITIALIZED
- This changes the test so image is cratead with valid layout, than transitioned to COLOR_ATTACHMENT_OPTIMAL using additional barrier
See merge request !207
Pyry Haulos [Sun, 27 Dec 2015 03:25:23 +0000 (22:25 -0500)]
Merge branch '117-push-constant-wrong-glsl' into 'master'
Fix bugs of the shaders in push constant tests
This fix issue #117 and #118
-fix the offset of the uniform block member to align with it's size
-modify the shader to fix wrong expected value for dEQP-VK.pipeline.push_constant.graphics_pipeline.data_update_partial_1
See merge request !199
Pyry Haulos [Sun, 27 Dec 2015 03:24:28 +0000 (22:24 -0500)]
Merge branch 'jekstrand-fix-timestamp-c++11' into 'master'
PipelineTimestampTests: Stop relying on C++11 features
The cbegin and cend functions weren't added to std::vector until c++11.
They also aren't really serving a purpose here given that there's no
automatic type inference going on, just a regular for loop. This fixes the
build on Fedora 23 when using clang and default dEQP build flags.
See merge request !198
Pyry Haulos [Sun, 27 Dec 2015 03:23:55 +0000 (22:23 -0500)]
Merge branch '76-fix-commandbuffer-params' into 'master'
fix VkCommandBufferBeginInfo struct values, fixes #76
just a short fix to change those two mixed struct values
See merge request !190
Pyry Haulos [Sun, 27 Dec 2015 03:22:03 +0000 (22:22 -0500)]
Merge branch 'jbolz_fix_ubo' into 'master'
Fix ubo.per_block_buffer tests
See merge request !187
Pyry Haulos [Sun, 27 Dec 2015 03:20:59 +0000 (22:20 -0500)]
Merge branch 'jekstrand-binding-model-hang-fix' into 'master'
api.binding_model: Wait for idle before throwing NotSupportedError
If we don't, then the GPU could be busy and some of the objects destroyed
by the exception propagation could be in-use. This is something of a
stop-gap measure; the problem of exceptions and the GPU being busy is
something that needs to be solved in general. However, it does fix GPU
hangs on some drivers.
This is related to #20.
See merge request !185
Jason Ekstrand [Fri, 18 Dec 2015 18:09:41 +0000 (10:09 -0800)]
api.binding_model: Wait for idle before throwing NotSupportedError
If we don't, then the GPU could be busy and some of the objects destroyed
by the exception propagation could be in-use. This is something of a
stop-gap measure; the problem of exceptions and the GPU being busy is
something that needs to be solved in general. However, it does fix GPU
hangs on some drivers.
This is related to #20.
Rocky Shi [Thu, 24 Dec 2015 04:13:43 +0000 (12:13 +0800)]
Fix typo in compute shader and assertion when creating buffers.
Jeff Bolz [Thu, 24 Dec 2015 00:09:26 +0000 (00:09 +0000)]
Fix number of command buffers for secondary_execut_twice test. Fix result buffer memory binding for order_bind_pipeline test. Fix comments for CommandBufferAllocateInfo structs.
Daniel Koch [Wed, 23 Dec 2015 19:42:40 +0000 (14:42 -0500)]
Move draw shaders to correct location
The draw/ tests are expecting the path vulkan/draw/ instead of just draw/.
Kristof Kosztyo [Fri, 18 Dec 2015 14:34:13 +0000 (15:34 +0100)]
Style fixes for the Copy and blitting