* Adding remaining flavors for grouped conv fwd
As titled. Following variants are added:
- grouped_conv2d_fwd_dynamic_op
- grouped_conv3d_fwd_dynamic_op
- grouped_conv3d_fwd_bilinear
- grouped_conv3d_fwd_convscale
- grouped_conv3d_fwd_convinvscale
- grouped_conv3d_fwd_convscale_add
- grouped_conv3d_fwd_convscale_relu
- grouped_conv3d_fwd_scale
- grouped_conv3d_fwd_combconvscale
- grouped_conv3d_fwd_scaleadd_scaleadd_relu
* Fix incomplete parsing of types from source names in add_instance_library() cmakelists function so we don't build f8 on RDNA3.
* Do not build f8 / bf8 only flavor tests on RDNA3
* Make sure we have proper generic instances for all instance lists related to the post-ces extra flavors, with scalarPerVector = 1. Then disable all but one generic instance per instance list to reduce compile time.
* Post rebase fix: Template parameters for Grouped Conv Fwd Device Impl got tweaked upstream.
* adding int8 and fp16 overloads to the elementwise operations
* fixed copilot nits
* Addressing review comments:
- removed unnecessary examples for dynamic op
- removed unnecessary conv specalizations for all the flavors
- removed spurious bilinear and scale source files
* clang-format
* reduced no of tests
---------
Co-authored-by: Wojciech Laskowski <wojciech.laskowski@streamhpc.com>
* chore: split block scale example instances in more separate files to speed up compile times
* wip: fp4 scaffolding for abquant
* feat: add fp4 decoding-while-loading to abquant pipeline
* feat: add support for fp4 CPU verification in abquant
* chore: add time tracking to reference calculation
* feat: add a4w4 test for blockscale gemm
* feat: optimize reference calculation by preconverting values to AccType
* feat: add fp4 to fp8 look-up table
* fix: reference to wrong ComputeDataType field in QuantProblem
* feat: type utilities for determining MFMA compute types
* feat: packed fp4 for abquant weight preshuffle
* feat: add separate tests for a4w4 base case, padding and preshuffleB
* fix: fp4 conversion on gfx950 attempting to use non-supported method
* fix: test case was using quant group sizes which don't work on gfx950 due to larger mfma tile size
* chore: add fp4 preshuffleb mode to block scale example
* chore: sanity check for packed types being 1 byte
* chore: clarify tensor dimension indices with constants
* chore: replace traits check with specialized check for packed types
* style: some minor refactoring and cleanup
* fix: correct conversion table for FNUZ fp8
* chore: add fp4 instances to main abquant instances again
* chore: use same initialization branch for int4 and fp4
* chore: add missing initialization for fp4 in block scale gemm example
---------
Co-authored-by: Thomas Ning <Thomas.Ning@amd.com>
* Decouple configure/build/test tools from Docker
Create a two-layer tool architecture:
- Core tools (ck-configure, ck-build, ck-test): Environment-agnostic,
work on any system with ROCm - no Docker dependency
- Container tools (ck-docker): Manage Docker containers and delegate
to core tools via docker exec
Changes:
- Add ck-configure: New CMake configuration tool with preset support,
native GPU detection, and flexible options
- Refactor ck-build: Remove Docker dependency, add --configure and
--list options, call ninja directly
- Refactor ck-test: Remove Docker dependency, add CTest integration
with --smoke/--regression/--all options
- Enhance common.sh: Add native GPU detection, build directory utils,
and output helpers
- Update ck-docker: Add configure/build/test/exec commands that
delegate to core tools inside container
This enables:
- Native development on ROCm hosts without Docker
- Simpler CI/CD integration
- Consistent behavior inside and outside containers
Co-Authored-By: Claude <noreply@anthropic.com>
* Add ck-rocprof: GPU profiling tool for rocprof-compute
Adds a command-line profiling tool to simplify GPU performance
analysis workflow using AMD rocprof-compute.
Features:
- Easy setup with automatic Python venv configuration
- Simple CLI: setup, run, analyze, compare, list
- Automatic GPU architecture detection
- Focus on LDS metrics (Block 12) for bank conflict analysis
- Comprehensive documentation with examples and troubleshooting
Usage:
ck-rocprof setup # One-time environment setup
ck-rocprof run <name> <executable> # Profile executable
ck-rocprof analyze <name> [block] # Analyze metrics
ck-rocprof compare <name1> <name2> # Compare two runs
ck-rocprof list # List available runs
* Make ck-rocprof documentation concise and improve Docker integration
- Streamlined documentation from 416 to 157 lines (62% reduction)
- Focused on essential commands, metrics, and workflows
- Enhanced script to run all operations inside Docker containers
- Fixed workload directory path and improved container management
- Added automatic rocprofiler-compute installation and dependency handling
* Add --no-roof flag to ck-rocprof profile command
Skip roofline analysis by default to speed up profiling. Roofline
analysis can add significant time to profiling runs but is not
needed for most LDS bank conflict analysis workflows.
* Make ck-rocprof work independently of Docker
Add native execution mode that runs rocprof-compute directly on the host
system when available, falling back to Docker mode when not.
Key changes:
- Auto-detect native mode when rocprof-compute is in PATH or common locations
- Add execution mode wrappers (exec_cmd, file_exists, dir_exists, etc.)
- Native mode stores venv at .ck-rocprof-venv in project root
- Native mode stores workloads at build/workloads/
- Support user-installed rocprofiler-compute (e.g., ~/.local/rocprofiler-compute)
- Add CK_FORCE_DOCKER env var to force Docker mode
- Update help message to show current execution mode
- Maintain full backward compatibility with existing Docker workflow
Tested successfully with rocprofiler-compute 3.4.0 installed from source
on MI300X GPU in native mode.
Co-Authored-By: Claude <noreply@anthropic.com>
* Add clean/status commands and improve ck-rocprof robustness
- Add 'clean' command to remove profiling runs (supports --all)
- Add 'status' command to show configuration and environment info
- Add workload name validation to prevent path traversal attacks
- Fix uv installation to use pip instead of curl for reliability
- Add cross-platform stat support for macOS compatibility
- Consolidate ROCPROF_CANDIDATES to avoid code duplication
- Expand help documentation with all profiling block descriptions
- Fix Docker wrapper script escaping issues
Co-Authored-By: Claude <noreply@anthropic.com>
* Fix analyze command to use correct workload path
rocprof-compute stores results directly in the workload directory
(pmc_perf.csv) rather than in a GPU architecture subdirectory.
Updated find_workload_path to detect this correctly.
Co-Authored-By: Claude <noreply@anthropic.com>
* Address PR review security and robustness issues
Security fixes:
- Escape executable path in cmd_run to prevent shell injection
- Add workload name validation to cmd_analyze and cmd_compare
Robustness improvements:
- Add error checking for uv package manager installation
- Use consistent project root detection (find_project_root || get_project_root)
- Use /opt/rocm instead of hardcoded /opt/rocm-7.0.1 in Docker mode
- Derive ROCM_REQUIREMENTS path from ROCPROF_BIN for flexibility
- Use gfx950 as fallback GPU consistent with common.sh
Documentation updates:
- Fix env var name GPU_TARGET -> CK_GPU_TARGET
- Update storage layout to reflect current structure (workloads/<name>/)
- Document clean and status commands
- Clarify native vs Docker default paths
Co-Authored-By: Claude <noreply@anthropic.com>
* Simplify ck-rocprof to native-only mode
Remove Docker mode from ck-rocprof. Docker users should run the tool
via `ck-docker exec ck-rocprof ...` instead.
This simplification:
- Removes ~210 lines of Docker-specific code
- Eliminates mode detection complexity
- Makes the script easier to maintain
- Provides clearer error messages when rocprof-compute is not found
The setup command now lists all searched locations when rocprof-compute
is not found, helping users understand how to install it.
Co-Authored-By: Claude <noreply@anthropic.com>
* Add rocprofiler-compute source installation fallback
When rocprof-compute is not found in system locations, automatically
install rocprofiler-compute 3.4.0 from source as a fallback. This
eliminates the hard dependency on system ROCm packages.
Implementation details:
- Clone rocprofiler-compute from GitHub to ~/.local/
- Install dependencies via requirements.txt (not editable install)
- Create wrapper that sets PYTHONPATH to source directory
- Execute source script directly rather than importing as module
This approach matches the project's development workflow and works
around the incomplete pyproject.toml that prevents editable installs.
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* create a filter to build only libs required by hiptensor
* allow building libs for miopen and hiptensor at the same time
* tweak the lib filtering logic one more time
* Add multi AB support to wave transfer
* Improviments to multi ABD examples
* Add instances and use intrawave v1 instead of interwave
* Apply changes to other transfers
* Wave transfer: add support for multiple internal vgpr buffers
* Fix compilation error gfx11
* Enable bwd weight splitk autodeduction with cap
* Fix error threshold calculations
* Add missing logic to wmma multiple d kernel
* Fix threshold calculation
* Update test with new applicability
Force merging because I verified this fix manually:
git checkout develop
git pull
ninja smoke-builder (failed to build, as expected)
git checkout rvoetter/ckb-fix
ninja smoke-builder (passed!)
* initial commit
* preshuffleQuant support for ABQuant
* fix mxfp4 to use correct QuantGroupSize
* addressing review comments and seperated Preshufflequant for A and B
* updated grouped gemm example for updated traits definition
* fix for CI failure
* updated grouped_gemm_abquant test for updated traits definition
* updated grouped_gemm_abquant test for updated traits definition
- Add multi-dimensional page index support (YsGatherDims) in tile_scatter_gather
- Add is_gather_dim() and get_gather_index() for multi-dim page lookup
- Override MakeVDramTileDistribution() for VECTORIZED_LAYOUT to match
GEMM's BWarpDstrEncoding (K decomposition: {K2, K0, K1})
- Add GetGemmKDecomposition() to retrieve kABKLane and kKPerThread
- Add static_assert for RowMajor VLayout requirement in batch prefill
Co-authored-by: Po Yen Chen <PoYen.Chen@amd.com>
* ck-builder: tensor copy function
This function copies one tensor to another, so that the memory
layout can be changed between them.
* ck-builder: fix ck::bhalf literals
These types don't work properly.
* ck-builder: abstract compare_elements in gpu_verification.hpp and make builder use it
This reduces the amount of duplicated code a bit.
* ck-builder: add flat tensor iterator
This "iterator" type pretends to be a pointer, useful for passing
tensors to functions expecting pointer-like types.
* ck-builder: integrate validation with ck gpu verification
By templating the gpu_verify function over iterators, we can use
the new FlatTensorIterator to adapt the function to multi-
dimensional tensors without changing either implementation
too much.
* ck-builder: add check_by_accumulations
This changes the gpu_verification.hpp code to also accept "iterator"
types for the relevant gpu_verify and gpu_reduce_max functions.
* ck: fix test_gpu_verification GenerateRandomData for bhalf
is_integer_it<bhalf_t> yields true, but it is not actually
an integer.
* ck: make gpu_verification kernels be proper persistent kernels
Previously these were using a hardcoded value for the grid size. This
commit changes that so that the grid size is automatically derived
from the kernel's occupancy and the number of multiprocessors on
the GPU.
* ck: clean up gpu_verification.hpp using block_reduce
This implements a small generic block reduce function, and rewrites
the rest of gpu_verification.hpp using that function to clean it up
a bit.
* ck-builder: doc typos
* ck-builder: update testing readme with validation interface.
* ck-builder: rebase fixes + review comments
* ck-builder: fix device integer generation with float types
Passing bfloat here causes a nans due to type_convert performing
a bitcast.
* ck: another bhalf_t bug
CK expects that int-generation with ck::bhalf_t yields bhalf integers,
not unsigned integers. This makes the logic of FillUniformRandInteger
compatible with GeneratorTensor_2<InDataType>, however idiotic that
may be.
* added reflection for conv_fwd_multiple_d_wmma_cshuffle.hpp
* added reflection for device_grouped_conv_bwd_weight_xdl_cshuffle
* added reflection for device_grouped_conv_bwd_weight_xdl_cshuffle v3
* added reflection of max_transpose parameters
* fix printing of std optional parameters
* fix use of undefined ck::index
* added conv traits for device_grouped_conv_bwd_weight_multiple_d_xdl_cshuffle
* added xdl two stage instance to reflection
* added additional variables
* added reflection for grouped_conv_bwd_weight_multiple_d_wmma_cshuffle, _v3, grouped_conv_two_stage_wmma_cshuffle_v3,
* added reflection for device_grouped_conv_bwd_weigh_wmma_cshuffle_v3
* added reflection for bwd_weight_wmma_cshuffle
* added comments back in
* add printed output for optional parameters
* update README
* fix typo
* added num_gemm_k_prefetch_stage and small fixes
* modified test string due to reflection of new parameter
---------
Co-authored-by: Kevin Abraham <kevin.abraham@streamhpc.com>
1. Add base class GridwiseGemm_xdl_cshuffle_base for all gridwise_gemm_xdl classes.
- to select correct LDS layout and epilogue behavior , three additional parameters is added.
- ForceNaiveLdsLayout: disable XOR based LDS layout when it is true
- DirectLoad: pipeline only use directload, we need force naive layout and ignore any padding on gfx9
- IsMxGemm: epilogue has two addtional dimensions
2. Move all LDS descriptor layout related fucntion to base class, including
- GetABlockDescriptor_AK0PerBlock_MPerBlock_AK1
- GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1
- GetCShuffleBlockDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
3. Move several LDS related helper funtions to base class, including
- GetSharedMemoryNumberOfByte
- GetABlockDescriptor_AKB_AK0PerBlock_MPerBlock_AK1
- GetBBlockDescriptor_BKB_BK0PerBlock_NPerBlock_BK1
- GetCBlockDescriptor_MBlock_NXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl
4. Move all c epilogue related code to base class, and 4 kind of implementation are provided
- RunEpilogueNoShuffle
- RunEpilogue
- RunMultiDEpilogue
- RunMoeEpilogue
* removed the api reference
* updating to the latest rocm-docs-core min version
* fixed a formatting issue with buffer views
* removed reference links from code snippets
* removed reference links from code snippets
---------
Co-authored-by: John Afaganis <john.afaganis@amd.com>
This document describes techniques for reducing C++ template instantiation
overhead in the Composable Kernel codebase, including:
- Replacing recursive templates with pack expansion (O(N) → O(1) depth)
- Using named functors instead of lambdas to share instantiations
- Replacing template recursion with constexpr loops
- Using fold expressions for accumulation operations
These techniques can significantly reduce build times for template-heavy code.
* ck-builder: restructure testing conv
In order to prepare for bwd of conv testing, this commit moves some
files and types around so that we can reuse ckt::Args for both forward
and backwards convolution.
* ck-builder: decouple fwd_ck.hpp and fwd_reference.hpp from fwd.hpp
This will allow us to more easily include fwd.hpp from backwards
definitions, which is required for initializing bwd values.
* ck-builder: fix layout of test_ckb_conv_bwd_weight_xdl_cshuffle_v3
Turns out that the supplied layout isn't actually supported...
* ck-builder: ck and reference conv integration for bwd weight
* ck-builder: ck bwd weight execution test
* ck-builder: ckt::run support for ck-tile bwd weight
* ck-builder: ck tile bwd weight execution test
* ck-builder: extra debug printing in MatchesReference
* ck-builder: make ckt::run return RunResult
This type is more convenient than std::tuple, as it will allow us to
use google test matchers with this in the future.
* ck-builder: RunResult matcher
Using EXPECT_THAT(..., SuccessfulRun()) will generate a check and a nice error
message about how and why running an algorithm failed.
* ck-builder: doc fixes
* ck-builder: add missing headers
This PR introduces a Python toolkit for analyzing Clang's `-ftime-trace` build performance data. This is the foundation for our systematic effort to reduce CK and CK-Tile build times (#3575).
The toolkit provides fast parsing of trace JSON files into pandas DataFrames using orjson, with specialized functions for analyzing template instantiation costs and compilation phase breakdowns. It includes a core library (`trace_analysis/`), example scripts for quick analysis, a comprehensive README with usage documentation, and an interactive Jupyter notebook demonstration.
Key features include memory-efficient DataFrame schemas with optimized dtypes, recursive hierarchical phase analysis, automatic metadata extraction (source file, compilation timing), and template instantiation filtering. The design supports both standalone scripts and interactive Jupyter notebook workflows.
This single-file analysis capability lays the groundwork for future multi-file analysis across thousands of compilation units, enabling data-driven optimization and build time regression detection.
* Add padding support with transpose
Also move check before writing storing is_src_valid during reading
* Add/modify instances to use wave transfer for gemm universal
Condition is changed so now the vectorsize of vmem reading and lds
writing must be equal to 8 in order to use the wave transfer
* Fix clang format
* Modify example
* Fix bwd data
* Add restriction for wave transfer with padding and transpose
Add test case which shows this limitation
* Fix validity checks 8 bit types
* Add validity check gemm_bias_add_reduce
* Add validity check grouped gemm tile loop
* Fix validity checks new flavours
* Minor fixes
* Fix clang format
* WIP: host level interwave pipeline compiles
* WIP: interwave implementation computes correct GEMM result when no aquant
* WIP: quantization works for subset of problem shapes
* WIP: quantization works for subset of problem shapes
* WIP: interwave memory pipeline passes local test
* feat: Add interwave pipeline implementation for memory pipline in aquant
* test: add unit test for aquant memory pipeline
* WIP: host level interwave pipeline compiles
* WIP: interwave implementation computes correct GEMM result when no aquant
* WIP: quantization works for subset of problem shapes
* WIP: quantization works for subset of problem shapes
* WIP: interwave memory pipeline passes local test
* feat: Add interwave pipeline implementation for memory pipline in aquant
* fix: compilation error on gfx950
* chore: remove debug statements from the code
* test: resolve merge conflict
* test: remove non rcr unit tests from test suite
* Re-enable f8 x bf8 tests on CompV3 as they now pass
* On CompV4, fp8 x bf8 tests now pass with K_BlockSize I32
* Add a changelog entry
---------
Co-authored-by: Illia Silin <98187287+illsilin@users.noreply.github.com>
* Added common struct to enable code reduction in gemm gemm and gemm multi_d gemm multi_d wmma implementation
This file includes all shared components. The (shared between the two implementations) kernel, the pointer offset computation struct, the grid descriptor creator and definitions, the invoker struct and the argument struct.
Signed-off-by: Chris Tsiaousis <chris.tsiaousis@streamhpc.com>
* Used the common struct in the batched gemm gemm wmma cshuffle v3 implementation
Signed-off-by: Chris Tsiaousis <chris.tsiaousis@streamhpc.com>
* Used the shared structs in the gemm multiple D gemm multiple D wmma cshuffle v3 implementation
Signed-off-by: Chris Tsiaousis <chris.tsiaousis@streamhpc.com>
* Boy-scout: IWYU paradigm in the gemm gemm and gemm multiple D gemm multiple D wmma cshuffle v3 implementations
Signed-off-by: Chris Tsiaousis <chris.tsiaousis@streamhpc.com>
---------
Signed-off-by: Chris Tsiaousis <chris.tsiaousis@streamhpc.com>
This change significantly improves compile-time performance by reducing template
instantiation depth for sequence generation and merging operations:
Optimizations:
- sequence_gen: Reduce instantiation depth from O(log N) to O(1) by using
__make_integer_seq to generate indices in a single step, then applying the
functor via pack expansion
- uniform_sequence_gen: Similarly optimized to O(1) depth using __make_integer_seq
with a helper that applies a constant value via pack expansion
- sequence_merge: Reduce depth from O(N) to O(log N) using binary tree reduction
strategy. Added direct concatenation specializations for 1-4 sequences to
avoid recursion in common cases, falling back to binary tree merging for 5+
sequences
Documentation:
- Added extensive inline comments explaining why sequence_merge cannot achieve
O(1) depth like sequence_gen (requires computing cumulative sequence lengths
from heterogeneous inputs, inherently requiring recursion)
- Documented the binary tree reduction approach and why it's superior to fold
expressions for this use case
Testing:
- Added comprehensive unit tests for uniform_sequence_gen with different values,
sizes, and edge cases
- Added tests for sequence_gen with custom functors (double, square, identity,
constant) to verify the new implementation works with arbitrary functors
- Added tests for sequence_merge with 4, 5, and many sequences to verify both
the direct concatenation path and binary tree reduction path
- Added tests for empty sequence edge cases
* Fix alignment issue in Stream-K workspace buffer
In CK Tile Stream-K, the workspace buffer is used to hold flags and
partials, where the first i bytes holds the flags and the remaining
bytes hold partials. This change adds padding to the flags prefix of the
workspace buffer to ensure the number of bytes is 128B-aligned. Without
this alignment, since workgroups do not skip cache when reading from
partials, they may read stale partials data in cache, leading to
incorrect results. The added padding avoids the stale data reading.
This change also re-enables the test_ck_tile_streamk_reduction tests.
* Compute reference GEMM on GPU for test verification to decrease testing time
* Moved device struct for batched gemm wmma to a common file
Signed-off-by: Chris Tsiaousis <chris.tsiaousis@streamhpc.com>
* Use the common device struct in the scaled batched gemm wmma implementation
Signed-off-by: Chris Tsiaousis <chris.tsiaousis@streamhpc.com>
* Boy-scout: Remove unused includes and ambiguous comment
Signed-off-by: Chris Tsiaousis <chris.tsiaousis@streamhpc.com>
* Moved pointer offset calculation and gridwise argument to common struct
This change enables further code reduction by re-using the common structs for the batched gemm and batched gemm b scale wmma implementations.
Signed-off-by: Chris Tsiaousis <chris.tsiaousis@streamhpc.com>
* Moved type string to the common struct of DeviceBatchedGemm_Wmma_CShuffleV3_Common"
Signed-off-by: Chris Tsiaousis <chris.tsiaousis@streamhpc.com>
---------
Signed-off-by: Chris Tsiaousis <chris.tsiaousis@streamhpc.com>
* Added bias_bnorm_clamp for WMMA conv fwd large tensor.
Following operations are added for FP16/BF16 data type and NHWGCxGKYXC layout.
- grouped_conv2d_fwd_bias_bnorm_clamp
- grouped_conv3d_fwd_bias_bnorm_clamp
* changed strategy to handle GemmArgs array
* Adding generic instance
* fixed last nits from reviewers and copilot
