## Motivation
PR #4797 added CK Tile bwd weight kernels to the CK Profiler. The
two-stage kernels were not supported in the initial PR. This PR adds the
the missing bwd weight two-stage kernels to the CK Profiler.
## Technical Details
Extended the CK Tile conv builder factory to build also the elementwise
ops required for the two-stage kernels. Extended the CK Builder for CK
Tile instance to accept the two-stage flag as part of the algorithm
configuration.
## Test Plan
Added units tests for CK Builder that verify the two-stage kernel
construction.
## Test Result
If CI passes, the added unit tests are passing.
## Submission Checklist
- [x] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
---------
Co-authored-by: Ville Pietilä <>
Add factory, InstanceTraits, and conv traits support for the WMMA V3
forward convolution kernel, enabling the CK Builder to generate and
dispatch this kernel variant used by MIOpen on gfx11/gfx12 GPUs.
## Motivation
As reported in issue #4944, MIOpen includes WMMA V3 forward convolution
kernels, so this PR adds support for those kernels similarly to other
supported kernels.
## Technical Details
This follows the same implementation as the other kernels. I added some
support for reflection, but I left a few todos since we need to
generalize our convolution traits to generalize across WMMA/MFMA and
CK/CKTile.
## Test Plan
Added faster tests to `ninja smoke-builder` that check the
instance-traits logic, and I added longer tests that instantiate
kernels, following the existing pattern in other kernals.
## Test Result
I tested all code with `ninja check-builder` on a gfx1101 build and ran
on gfx1101.
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
## Motivation
This PR fixes how CK Builder is validating transfer vector size and adds
proper validation for LDS transfer vector size as well.
## Changes:
* [__source vector dim__] -- Before this PR the data transfer validation
logic didn't allow to set the source vectorized dimension to 1. However
there are CK instances that are doing this when the group merging is
used. This is used only for
`DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle` kernel.
* [__valid vector size__] -- Before this PR the validation logic
concerned only single instruction maximum vector size. However our
buffer loading logic has implemented support for loading more values
through multiple buffer instructions. This again was discovered to be
used in some of the convolution instances. Thus this behavior was
reflected in validation logic.
* [__valid LDS vector size__] -- Before this PR the LDS vector size
validation was done in the same way as VMEM. This PR adds proper LDS
vector size validation based on the available LDS instruction sizes.
## Test Plan
Run CK BUILDER conv fwd factories tests
## Test Result
All CK BUILDER conv fwd factories work (except DL one & ck tile since
they're not yet added now)
## Submission Checklist
- [ ] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
This PR adds bwd data wmma and xdl kernels to the ck builder, their
instance and conv traits as well as tests for the above.
---------
Co-authored-by: Kevin Abraham <kevin.abraham@streamhpc.com>
Co-authored-by: John Shumway <jshumway@amd.com>
* Rename member variable to better reflect its actuall meaning.
* Add transfer checks for conv fwd xdl.
* Validate tensor layouts & vector size conv fwd v3.
* Add combined transfer concepts.
* Add transfer concepts for conv fwd factories.
* Fix clang format
* Add helper instruction to get max mem vector instruction width.
* Apply review comments.
* Rename thread cluster access(->arrange) order concept
* FIx merge artifacts.
* Add generic access order limits into block transfer concept.
[ROCm/composable_kernel commit: 1a6d1b59ef]
* Add placeholder test.
* Initial conv bwd weight factory.
* Conv builder test refactoring.
* Add missing pieces to bwd weight factory.
* Improve compile time erros message when no matching factory is found.
* Use amcro to ensure automatic macthing between concepts are their string representations.
* Improve compile time diagnostics.
* Small improvements.
* Improve missing member/wrong type compile-time errors.
* Improve compile time diagnostics.
* Concept bug fixes.
* Remove debug assert.
* Update algorithm signature diagnostics.
* Factory bug fixes.
* First functional version of bwd weight conv factory.
* Refactor handing of GEMM-K batch template parameter in conv bwd weight factory.
* Concept improvements.
* Improve concept diagnostics.
* Introduve a common size type for concepts.
* Update compiletime diagnostics to use the size type.
* Update conv specialization enum.
* Fix fwd conv builder tests.
* Fix smoke tests.
* Separate bwd weigth and bwd data tests into separate targets.
* Clean-up CK Tile builder tests.
* Add bwd weight XDL CShuffle V3 factory.
* Build conv bwd weigth v3 instances successfully.
* Add instance traits for DeviceGroupedConvBwdWeight_Xdl_CShuffleV3.
* Test fix.
* Add instance traits for bwd weight algorithms.
* Add unit tests for instance strings.
* Build new instance traits unit tests but exclude WMMA for now.
* Added factory for DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle.
* Conv bwd weight DL factory.
* Final implementation for bwd weight DL factory.
* Add test for creating DeviceGroupedConvBwdWeightMultipleD_Xdl_CShuffle instance.
* Add factory for DeviceGroupedConvBwdWeightMultipleD_Xdl_CShuffle
* Treat ref algorithm the same way as real algorithms in the dispatcher.
* Refactor large tensor support and WMMA configuration.
* Add factory and tests for DeviceGroupedConvBwdWeight_Wmma_CShuffleV3.
* Update Readme.
* Fix WMMA bwd weight tests.
* Added factory and tests for DeviceGroupedConvBwdWeightTwoStage_Wmma_CShuffleV3.
* Factory and tests for DeviceGroupedConvBwdWeight_Wmma_CShuffle.
* Dispatching for DeviceGroupedConvBwdWeightMultipleD_Wmma_CShuffle.
* Add factory for DeviceGroupedConvBwdWeightMultipleD_Wmma_CShuffleV3
* Fix DeviceGroupedConvBwdWeightMultipleD_Wmma_CShuffleV3 factory and compute types for input and output tensor in bwd weigth convs.
* Fix fwd factories after refactoring.
* clang-format
* Move compile-time diagnostics to a separate branch.
* Fix ref algorithm dispatching.
* Fix smoke tests.
* clang-format
* Fix factory for regular WMMA conv bwd weight.
* Clarify builder Readme.
* Remove obsolete test file.
* Fix test after merge.
* clang-format
* Remove the C++26 extensions.
* Unify conv elementwise ops and layout definitions for fwd and bwd directions.
* Remove old layout and elementwise ops.
* Unify handling of conv tensor types between fwd and bwd directions.
* Unify block transfer for fwd and bwd directions. Rename ThreadSliceDim to ThreadClusterRank.
* Make BlockTransferDescriptor concept parametrized. Introduce a common TileTransferParameters concept for conv algorithms.
* clang-format
---------
Co-authored-by: Ville Pietilä <>
[ROCm/composable_kernel commit: 9908a87c31]
* [CK_BUILDER] Integrate GPU reference as ConvAlgorithm
Add GPU reference as a ConvAlgorithm specialization, enabling:
- Unified Builder API for reference and optimized kernels
- Future ckProfiler integration for validation
- First step toward numerical validation in Builder tests
Changes:
- Add ConvAlgorithmSpecialization::REFERENCE enum
- Add ConvAlgorithm_Reference struct
- Add IsReferenceAlgorithm concept
- Create 3 reference factories (Forward, BwdData, BwdWeight)
- Wire into conv_dispatcher
- Add proof-of-concept test (passing)
Test result: Can instantiate reference through Builder API
* Add GPU reference execution tests
- Reference kernel executes through Builder (459ms)
- Both reference and optimized can instantiate
- Tests passing
Next: Implement utilities for comparison
* Optimized Builder kernel execution works
- MakeArgument pattern implemented
- Builder-generated kernel executes successfully
- Tests passing (451ms execution)
Next: Add comparison
* VALIDATION COMPLETE: Builder == Reference
Builder-generated kernel output matches GPU reference!
Test: Validate_Optimized_vs_Reference_Forward_2D_FP16
Result: PASS ✓
This proves CK Builder generates correct code!
* Update to new Builder API
All tests passing
* Rename test file for clarity
test_builder_kernel_execution -> test_builder_kernel_validation
* Add all 3 directions support
- Forward, Backward Data, Backward Weight
- All reference factories working
- Dispatcher wired for all directions
- 9 tests passing
Tests:
- test_reference_execution: 3 tests (all directions)
- test_optimized_execution: 3 tests (all directions)
- test_builder_kernel_validation: 3 tests (fwd validated, bwd placeholders)
* Add backward direction support
- Backward data and weight dispatcher wiring
- Fix factories for new API
- All 3 directions tested
- 9 tests passing
* Refactor: Change IsReferenceAlgorithm from concept to consteval function
Address review feedback: Use consteval function in dispatcher instead of
concept, matching the pattern for other algorithms (Tile, XDL, WMMA, DL).
- Remove IsReferenceAlgorithm concept from conv_algorithm_concepts.hpp
- Add IsReferenceAlgorithm() consteval function to conv_dispatcher.hpp
- Update dispatcher to use function call: IsReferenceAlgorithm<T>()
- Remove redundant algorithm checks from reference factory requires clauses
All tests passing (9/9).
* Move Tile algorithm check outside direction block to support all directions
* Implement MakeInvokerPointer interface and add random input validation
- Implement full Argument/Invoker structs for old CK interface (not just nullptr)
- Refactor with reference_common.hpp to reduce code duplication
- Add random input validation tests: Builder vs direct GPU reference (all directions)
- Fix layout: GNHWC -> NHWGC to match reference kernel expectations
- All 12 tests pass with IDENTICAL results on random input
* Move ConvAlgorithm_Reference to test/impl/conv_algorithm_types.hpp
Keep types.hpp for data types only (enums), move algorithm descriptors
to conv_algorithm_types.hpp as suggested by review.
* Add static_assert to ensure reference factories only accept PassThrough operations
Reference implementation doesn't support fused elementwise operations.
Add compile-time validation to fail early with clear error message if
non-PassThrough operations are specified on input, weight, or output.
* Add InstanceTraits support for reference kernels
- Store SIGNATURE/ALGORITHM/VERSION in Instance for reflection
- Create shared ReferenceCommonTraits base for common properties
- Add 3 direction-specific InstanceTraits specializations in one file
- Include data type and layouts in instance_string output
* Remove optimized kernel validation tests from reference-only branch
* Use existing layout helper and organize reference tests
Use LayoutToCK from conv_tensor_layout.hpp and move reference InstanceTraits
test to validation folder.
* Merge develop branch
Fix DataType switch for new mixed precision types.
* Fix comment spacing for CI
* Convert IsReferenceAlgorithm from function to concept
* Add reference tests to CI smoke tests
* Consolidate 3 reference factories into single unified factory
---------
Co-authored-by: Ville Pietilä <188998872+vpietila-amd@users.noreply.github.com>
[ROCm/composable_kernel commit: a0acc83a72]
* Separate layouts into separate entities for input, weight, and output tensors.
* Add test for handling bias tensor layouts.
* Use instance string in builder tests.
* Add handling of output bias data types and layouts.
* Generalize handling of the elementwise ops.
* Test fix.
* Create builder for layouts.
* Layout builder improvements.
* Improve layout builder.
* Simplify bias layout handling.
* Code clean-up.
* Move layout utils into separate file.
* Remove hard-coded layout combinations.
* Small code clean-up.
* Move data type utils into a separate file.
* Add data types, layouts, and elementwise ops per conv tensor.
* Builder bug fixes after refactoring.
* Working baseline.
* Make signature definition look nice in the test code.
* Move TensorConfig into test implementations.
* Fix all fwd conv builder tests.
* Fix conv traits and descriptors tests.
* More factory assets under a separate directory.
* Fix building conv traits.
* Fix clang-format.
* Add Readme doc to describe the design.
* Add link to main Readme. Fix links in the builder design doc.
* Clean-up data type/layout/elementwise op conversions.
* Switch from dimension and tensor type specific layouts to a flat list of tensor layouts.
* Fix clang-formatting.
* Fix clang-format for test code.
* Simplify fwd conv signature definitions in the test code.
* Remove accidental edits.
* Fix comment string.
* Fix instance factory after rebase.
* Fix tests after rebase.
* Unify layout handling.
* Add more conv layout unit tests.
* Clang-format.
* Fix merge conflicts.
* Improve elementwise op handling.
---------
Co-authored-by: Ville Pietilä <>
[ROCm/composable_kernel commit: 9cb1f421bc]
Refactor the builder factory code into multiple files and subdirectories and a ck_tile::builder::factory namespace.
The factory implements compile-time dispatch from high-level signature and algorithm descriptors to our existing specialized convolution kernel implementations.
Major changes in this PR:
Dispatch logic is explicit in the function make_conv_instance instead of implicit in template specialization selection.
Helper code is moved to a subdirectory builder/factory/helpers.
Helpers now have unit tests.
Factories are moved to their own files.
Code moved to namespaces ck_tile::builder::factory and ck_tile::builder::factory::internal.
This does not yet fix the problem of bad error messages, but the make_conv_instance function makes the poor error messages clear. The choice of algorithm must be much more robust (perhaps with explicit enumeration in the algorithm descriptor), so that the dispatch doesn't fail.
Quality changes:
Making dispatch explicit rather than implicit will improve robustness, readability, maintainability, testability, and extensibility.
Separating code into separate files and subdirectories helps readability and extensibility.
Adding unit tests for helpers documents behavior and will enable more complex logic and functionality.
Separating files (especially unit tests) helps clarify includes and dependencies and makes code easier to refactor.
[ROCm/composable_kernel commit: 280bc42191]
Our copyright were were mostly correct, but we inconsistently used (C) instead of (c) like the rest of the CK code. This PR fixes that (using lowercase c) and adds a missing copyright header to one file.
[ROCm/composable_kernel commit: f38c3de9f9]
Grouping transfer operations per tensor makes it easier to
constrain on and operate with the transfer operations. As an
example, we can now deduplicate the logic for translating
the transfer operations from the ck-builder interface to the old
ck interface for the A and B tensors.
[ROCm/composable_kernel commit: 245c6011cf]
Proposed changes
Improve the forward convolution builder implementation and addressed leftover feedback left from PR #3138. Main changes
Refactored tests such that they reflect better the builder pattern. The templates and types for the convolution algorithm concepts are created via factory that facilitates programmatic creation of the device op instances.
Moved tests into anonymous namespace.
The convolution factory had lot of if-else constructs when CK Builder types were converted into CK library types. I had initially trouble in using static_assert in the default branch of switch as the static_assert was evaluated at compile time even for valid types. However, if we change the static_assert to throw "<error message>", it will result in a compile-time error only if the default branch is actually hit. This assumes that the function is consteval. Hence, changed all conversions in the convolution factory to use switch, which is more intuitive.
Removed the explicit device op definition from convolution signature and the corresponding predicate file. The device ops are defined by the corresponding concepts. This allowed to remove lot of boilerplate code from the convolution factory.
Adde inheritance and convolution algorithm specialization to handle device ops that are specialization of a more generic ones. The large tensor support is more naturally expressed by this pattern.
Added support for the FP8 data type.
* WIP: Builder for expected test results.
* Improve ckb fwd conv instance tests.
* clang-format
* Change if-else statements into switch in conv factory.
* Fix clang-formatting.
* Removed unnecessary includes.
* Added missing copyright.
* Remove explicit device op flag from from convolution signature.
* Add missing concept.
* Fix build.
* clang-format
* Add test for building conv fwd FP8 instances.
* Add missing header to instance traits.
* Clean-up recently added instances.
* Introduce inheritance and specialization.
* Use builder to build conv algorithm templates and types.
* clang-format
* Fix conv description tests.
---------
Co-authored-by: John Shumway <john.shumwayjr@gmail.com>
[ROCm/composable_kernel commit: 7d57bc169f]
* Add device operation to conv signature. Use unions to hold conv layouts and device operations.
* Add predicates for all device op instances.
* Use the device op signature for validation.
* Fix ckb CMakeLists.txt file for tests.
* Fix building CK Builder instance traits after the introduction of direct load template parameter in CK.
* Fix clang-formatting.
* add device_grouped_conv_fwd_dl_multiple_d_nhwc_kyxc_nhwk
* Add full DL configurability with Option A implementation
- Added 5 DL descriptor structs (39 configurable parameters)
- Added 10 C++20 concepts for type-safe validation
- Updated factory to read all parameters from descriptors
- Updated test helper to populate all descriptors
- All tests passing (13/13 including 3 new DL tests)
* Add factory and test support for DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor
- Add factory specialization for Large_Tensor device operation (conv_factory.hpp lines 1145-1265)
- Add macro collision workaround using pragma push/pop (conv_factory.hpp lines 43-51)
- Add test helper function run_test_DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor
- Add builder test file test_ckb_conv_fwd_2d_large_tensor_fp16.cpp with 2 test cases
- Update CMakeLists.txt to include new test file
- Reuse existing ConvAlgorithm_DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle descriptor
- Map all 42 template parameters identical to regular XDL CShuffle
- All 15 builder tests passing including 2 new Large_Tensor tests
Completes Task 350: All 4 forward convolution device operations now supported in CK Builder.
* Update copyright headers to new format
- Change copyright format to: Copyright (C) Advanced Micro Devices, Inc., or its affiliates.
- Reorder headers: Copyright first, then SPDX-License-Identifier
- Updated files:
* experimental/builder/test/conv/test_ckb_conv_fwd_2d_dl_fp16.cpp
* experimental/builder/test/conv/test_ckb_conv_fwd_2d_large_tensor_fp16.cpp
* experimental/builder/include/ck_tile/builder/device_op_types.hpp
* fix c++ 18 format
* Fix clang-format-18 error in device_op_types.hpp
---------
Co-authored-by: Ville Pietilä <ville.pietila@amd.com>
Co-authored-by: Ville Pietilä <188998872+vpietila-amd@users.noreply.github.com>
[ROCm/composable_kernel commit: 5f3cae3e28]
* Update copyright messages.
Copyright messages should no longer include a year. This PR updates all 38 source files to the new format.
* Switch to (C) from unicode copyright symbol.
The unicodein comments was causing compilation errors.
[ROCm/composable_kernel commit: 0be0288f58]
* Add device operation to conv signature. Use unions to hold conv layouts and device operations.
* Add predicates for all device op instances.
* Use the device op signature for validation.
* Fix ckb CMakeLists.txt file for tests.
* Fix building CK Builder instance traits after the introduction of direct load template parameter in CK.
* Fix clang-formatting.
* Add factory for DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle device op.
* Add conv factory for DeviceGroupedConvFwdMultipleD_Wmma_CShuffle
* Rename elements per wave per shuffle member in the epilogue concept.
* clang-format
* Add concepts and types for optional device op template parameters.
* Add optional compute, direct load, and loop scheduler arguments to conv factory.
* Add number of groups to merge template parameter.
* clang-format.
[ROCm/composable_kernel commit: 3ae3992c18]
Generalize the current convolution factory in CK Builder to be able to build instances of any relevant convolution device operation. The main changes are:
* Added new enums FwdGroupConvDeviceOperation, BwdDataGroupConvDeviceOperation, and * BwdWeightGroupConvDeviceOperation that contain the device operations for which the builder should be able to build instances.
* Create a union structure GroupConvDeviceOp that can represent a single value of the fwd, bwd weight, or bwd data device operations. This would be more naturally represented by std::variant object, but we cannot use std::variant in NTTPs because it is not a structural object.
* Introduced a new member device_operation in the ConvSignatureDescriptor concept that assumes GroupConvDeviceOp value.
* Added predicates to be used in creation ConvFactory specialization for the different device operation. When we add support for a new device operation, we'll just create a new ConvFactory specialization with appropriate predicates.
* Changed handling of the convolution layouts (GroupConvLayout1D, GroupConvLayout2D, GroupConvLayout3D) to use the union based handling, i.e., there's now a GroupConvLayout union struct that can hold a single value of the 1D, 2D, or 3D layouts. This simplifies the handling of the different layouts as we get rid of templatized convolution signature.
These code changes allow developers to work more easily in parallel when adding new device operations.
* Fix building CK Builder instance traits after the introduction of direct load template parameter in CK.
* Fix clang-formatting.
[ROCm/composable_kernel commit: b387249fd9]
