Skip to content

Remove UNSUPPORTED labels from tests that are supported - #1487

Merged
bogner merged 6 commits into
llvm:mainfrom
bogner:2026-08-27-support-unsupported
Sep 3, 2026
Merged

bogner merged 6 commits into
llvm:mainfrom
bogner:2026-08-27-support-unsupported

Conversation

@bogner

@bogner bogner commented Aug 28, 2026

Copy link
Copy Markdown
Contributor

These all should be XFAILs, but they were marked as unsupported. Many of these pass now, so in those cases I've just removed them.

There are two more interesting cases:

@bogner bogner added the test-all When applied to a PR this will opt-in to additional pre-merge test configurations.. label Aug 28, 2026
Comment thread test/Feature/HLSLLib/countbits.16.test
Comment thread test/WaveOps/WaveActiveBitOr.int64.test
# Vulkan KosmicKrisp and MoltenVK the output is always False.
# UNSUPPORTED: Metal || KosmicKrisp || MoltenVK
# Note: For Vulkan KosmicKrisp and MoltenVK the output is always False.
# UNSUPPORTED: KosmicKrisp || MoltenVK

@farzonl farzonl Aug 28, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should we make this an XFAIL? If Metal is supported, then this would imply vulkan om Mac might eventually start working?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe. I'm not set up to test either of these and we don't have them in CI, so I was uncomfortable changing them.

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I removed all the KosmicKrisp || MoltenVK for the boolean matrix test and ran on both vulkan implementation and this issue appears to be resolved.

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Full Results:

KosmicKrisp Clang

********************
Failed Tests (11):
  OffloadTest-clang-vk :: Bugs/UAV-Sequental-Consistency.yaml
  OffloadTest-clang-vk :: Feature/HLSLLib/select.32.test
  OffloadTest-clang-vk :: Feature/HLSLLib/select.fp16.test
  OffloadTest-clang-vk :: Feature/MaximalReconvergence/loop_peeling.test
  OffloadTest-clang-vk :: Feature/ResourcesInStructs/res-of-matrix-in-struct.test
  OffloadTest-clang-vk :: Feature/Textures/Array.CalculateLevelOfDetail.test
  OffloadTest-clang-vk :: Feature/Textures/CalculateLevelOfDetail.test
  OffloadTest-clang-vk :: Feature/TypedBuffer/64bit-scalar.test
  OffloadTest-clang-vk :: Tools/Offloader/BufferFormats.test
  OffloadTest-clang-vk :: WaveOps/WaveActiveBitXor.int.test
  OffloadTest-clang-vk :: WaveOps/WaveActiveBitXor.int64.test

********************
Unexpectedly Passed Tests (2):
  OffloadTest-clang-vk :: Feature/HLSLLib/abs.long-vector.32.test
  OffloadTest-clang-vk :: WaveOps/GroupMemoryBarrierWithGroupSync.test

KosmicKrisp DXC

********************
Failed Tests (20):
  OffloadTest-vk :: Bugs/UAV-Sequental-Consistency.yaml
  OffloadTest-vk :: Feature/MaximalReconvergence/loop_peeling.test
  OffloadTest-vk :: Feature/ResourceArrays/overflow-unbounded-array.test
  OffloadTest-vk :: Feature/Semantics/CullDistance.test
  OffloadTest-vk :: Feature/Semantics/DomainSystemValues.test
  OffloadTest-vk :: Feature/Semantics/HullSystemValues.test
  OffloadTest-vk :: Feature/Textures/Array.CalculateLevelOfDetail.test
  OffloadTest-vk :: Feature/Textures/CalculateLevelOfDetail.test
  OffloadTest-vk :: Feature/TypedBuffer/64bit-scalar.test
  OffloadTest-vk :: Feature/Vk.SampledTextures/Vk.SampledTexture2D/Vk.SampledTexture2D.CalculateLevelOfDetail.test.yaml
  OffloadTest-vk :: Graphics/IsolineDomainTessellation.test
  OffloadTest-vk :: Graphics/QuadDomainTessellation.test
  OffloadTest-vk :: Graphics/gs_point_to_quad.test
  OffloadTest-vk :: Graphics/gs_selective_output.test
  OffloadTest-vk :: Tools/Offloader/BufferFormats.test
  OffloadTest-vk :: WaveOps/WaveActiveBitAnd.convergence.test
  OffloadTest-vk :: WaveOps/WaveActiveBitOr.convergence.test
  OffloadTest-vk :: WaveOps/WaveActiveBitXor.convergence.test
  OffloadTest-vk :: WaveOps/WaveActiveBitXor.int.test
  OffloadTest-vk :: WaveOps/WaveActiveBitXor.int64.test

MoltenVK Clang

********************
Failed Tests (20):
  OffloadTest-clang-vk :: Bugs/UAV-Sequental-Consistency.yaml
  OffloadTest-clang-vk :: Feature/HLSLLib/select.32.test
  OffloadTest-clang-vk :: Feature/HLSLLib/select.fp16.test
  OffloadTest-clang-vk :: Feature/MaximalReconvergence/loop_peeling.test
  OffloadTest-clang-vk :: Feature/ResourcesInStructs/res-of-matrix-in-struct.test
  OffloadTest-clang-vk :: Feature/Textures/Array.GatherCmp.test
  OffloadTest-clang-vk :: Feature/Textures/Array.SampleBias.test
  OffloadTest-clang-vk :: Feature/Textures/Array.SampleCmp.test
  OffloadTest-clang-vk :: Feature/Textures/GatherCmp.test
  OffloadTest-clang-vk :: Feature/Textures/Sample.test
  OffloadTest-clang-vk :: Feature/Textures/SampleBias.test
  OffloadTest-clang-vk :: Feature/Textures/SampleCmp.test
  OffloadTest-clang-vk :: Feature/TypedBuffer/64bit-scalar.test
  OffloadTest-clang-vk :: Tools/Offloader/BufferFormats.test
  OffloadTest-clang-vk :: WaveOps/QuadReadAcrossDiagonal.int64.test
  OffloadTest-clang-vk :: WaveOps/QuadReadAcrossX.int64.test
  OffloadTest-clang-vk :: WaveOps/QuadReadAcrossY.int64.test
  OffloadTest-clang-vk :: WaveOps/WaveActiveBitAnd.int64.test
  OffloadTest-clang-vk :: WaveOps/WaveActiveBitOr.int64.test
  OffloadTest-clang-vk :: WaveOps/WaveActiveBitXor.int64.test

********************
Unexpectedly Passed Tests (1):
  OffloadTest-clang-vk :: Feature/HLSLLib/abs.long-vector.32.test

MoltenVK DXC

Failed Tests (24):
  OffloadTest-vk :: Bugs/UAV-Sequental-Consistency.yaml
  OffloadTest-vk :: Feature/CBuffer/arrays-16bit.test
  OffloadTest-vk :: Feature/ConstantBufferT/nested.test
  OffloadTest-vk :: Feature/MaximalReconvergence/loop_peeling.test
  OffloadTest-vk :: Feature/Semantics/CullDistance.test
  OffloadTest-vk :: Feature/TypedBuffer/64bit-scalar.test
  OffloadTest-vk :: Graphics/IsolineDomainTessellation.test
  OffloadTest-vk :: Graphics/gs_point_to_quad.test
  OffloadTest-vk :: Graphics/gs_selective_output.test
  OffloadTest-vk :: Tools/Offloader/BufferFormats.test
  OffloadTest-vk :: WaveOps/QuadReadAcrossDiagonal.int64.test
  OffloadTest-vk :: WaveOps/QuadReadAcrossX.int64.test
  OffloadTest-vk :: WaveOps/QuadReadAcrossY.int64.test
  OffloadTest-vk :: WaveOps/WaveActiveAllEqual.int64.test
  OffloadTest-vk :: WaveOps/WaveActiveBitAnd.convergence.test
  OffloadTest-vk :: WaveOps/WaveActiveBitAnd.int64.test
  OffloadTest-vk :: WaveOps/WaveActiveBitOr.convergence.test
  OffloadTest-vk :: WaveOps/WaveActiveBitOr.int64.test
  OffloadTest-vk :: WaveOps/WaveActiveBitXor.convergence.test
  OffloadTest-vk :: WaveOps/WaveActiveBitXor.int64.test
  OffloadTest-vk :: WaveOps/WavePrefixProduct.int64.test
  OffloadTest-vk :: WaveOps/WavePrefixSum.32.test
  OffloadTest-vk :: WaveOps/WavePrefixSum.fp16.test
  OffloadTest-vk :: WaveOps/WavePrefixSum.int64.test

********************
Unexpectedly Passed Tests (3):
  OffloadTest-vk :: Feature/HLSLLib/cos.nan-inf-denorm.32.test
  OffloadTest-vk :: Feature/HLSLLib/sin.nan-inf-denorm.32.test
  OffloadTest-vk :: WaveOps/WaveActiveSum.int16.test


# UNSUPPORTED: Clang && Vulkan
# Clang-Vulkan is unsupported because of two validation errors
# This issue tracks its resolution: https://github.com/llvm/offload-test-suite/issues/285

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This issue is closed as not planned if support was added is that incorrect?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I changed it to closed as completed - the issue really didn't have enough information to determine what fixed it and I usually don't close things as completed unless I can verify that. It does seem that whatever was wrong is fixed now though, so sure.

# 16/64 bit firstbithigh doesn't have a DXC-Vulkan lowering
# Unsupported https://github.com/microsoft/DirectXShaderCompiler/blob/main/tools/clang/test/CodeGenSPIRV/intrinsics.firstbitlow.64bit.hlsl
# UNSUPPORTED: DXC && Vulkan
# XFAIL: DXC && Vulkan

@farzonl farzonl Aug 28, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should you add the DXC issue tracking this work here in a comment?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I pointed all of these at microsoft/DirectXShaderCompiler#8536, which technically only mentions 1 of the 4 cases but it seems like a reasonable enough place to track all 4.

Comment thread test/Feature/HLSLLib/firstbithigh.64.test
Comment thread test/Feature/HLSLLib/firstbitlow.16.test
Comment thread test/Feature/HLSLLib/firstbitlow.64.test

# DXC + Vulkan does not support multi-dimensional resource arrays
# UNSUPPORTED: DXC && Vulkan
# XFAIL: DXC && Vulkan

@farzonl farzonl Aug 28, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think if we are going to call these xfails now instead of unsupported we need an issue linked. If DXC was never going to get this feature, then unsupported seems to make more sense.

@bogner bogner Aug 28, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I do think if somebody implemented this we'd accept the patch, but unless someone comes along actually needing it I don't think it's worth filing an issue about. I've put it back to UNSUPPORTED.


# DXC + Vulkan does not support global structures containing buffers.
# UNSUPPORTED: DXC && Vulkan
# XFAIL: DXC && Vulkan

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same comment I think if we are going to call these xfails now instead of unsupported we need an issue linked. If DXC was never going to get this feature, then unsupported seems to make more sense.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same thing here - moved it back to UNSUPPORTED.

@bogner

bogner commented Aug 28, 2026

Copy link
Copy Markdown
Contributor Author

There are two new test failures that I need to track down or file issues for:

These all should be XFAILs, but they were marked as unsupported. Many of
these pass now, so in those cases I've just removed them.

There are two more interesting cases:

- `countbits.16` had a mistake in the test and is failing for clang. I
  filed llvm/llvm-project#219327 to track the clang failure.
- `SwitchBranchAttr` and `SwitchFlattenAttr` had the attribute in the
  wrong place and weren't actually testing what they said they were.
These don't have issues to associate with the XFAILs and we're extremely
unlikely to prioritize the work, so UNSUPPORTED is probably more appropriate.
@bogner
bogner force-pushed the 2026-08-27-support-unsupported branch from d0fee48 to 988e45f Compare September 2, 2026 19:34
Comment thread test/Basic/Matrix/matrix_bool_and_operator_thread_group.test Outdated
Comment thread test/Basic/Matrix/matrix_bool_or_operator.test Outdated
Comment thread test/Basic/Matrix/matrix_bool_and_operator.test Outdated

@bogner bogner left a comment

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

These tests pass on KosmicKrisp and MoltenVK

Comment thread test/Basic/Matrix/matrix_bool_and_operator.test Outdated
Comment thread test/Basic/Matrix/matrix_bool_and_operator_thread_group.test Outdated
Comment thread test/Basic/Matrix/matrix_bool_or_operator.test Outdated
Co-authored-by: Justin Bogner <mail@justinbogner.com>
@bogner

bogner commented Sep 3, 2026

Copy link
Copy Markdown
Contributor Author

Test failures are all unrelated matrix max/min tests that have been resolved upstream

@bogner
bogner merged commit 5e8c924 into llvm:main Sep 3, 2026
38 of 57 checks passed
Alexander-Johnston pushed a commit to Alexander-Johnston/offload-test-suite that referenced this pull request Sep 14, 2026
These all should be XFAILs, but they were marked as unsupported. Many of
these pass now, so in those cases I've just removed them.

There are two more interesting cases:
- `countbits.16` had a mistake in the test and is failing for clang. I
filed llvm/llvm-project#219327 to track the clang failure.
- `SwitchBranchAttr` and `SwitchFlattenAttr` had the attribute in the
wrong place and weren't actually testing what they said they were.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

test-all When applied to a PR this will opt-in to additional pre-merge test configurations..

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants