* Add reduction across all dims cases.
* host softmax: handle all reduce
* Test cases when reduced dim is not innermost axis.
* Fix syntax.
* Test non innermost dim for fp32 and int8
* Group test suites wrt NumReduceDim.
* Additionally test failing cases.
* Throw error when Rank or NumReduceDims doesn't match arguments.
* Check reducedDims has correct values
* Move don't reuse DeviceReduceMultiblock IsSupportedArgument method.
Instead implement own. (in fact just get rid of one check to enable
reduction across inner dimensions).
* Reorganize unit tests to better cover use scenarios.
* Test input validation
* Test reduction of inner dimensions with custom op instances.
* Refactor fp32 and int8 unit tests.
* Fix FP32 instance template parameters.
* Add more instances.
* Instances with InSrcVectorDim=0.
* Do not initialize and copy data when arg not supported.
* ckProfiler Softmax use instance factory.
* Refactor device softmax IsSupported.
* Additionally add non-polymorphic api functions
* Split softmax instances into multiple files.
* Fix profiler.
* Reorganize tests to reuse profiler and cover edge cases.
* Clang-format
* I8 Softmax instances along with UT.
* Reuse type alias definitions from instance factory header.
* Clean included headers
* Fix variable names.
* Add missing checks in Argument constructor.
Co-authored-by: Adam Osewski <aosewski@amd.com>
Co-authored-by: Anthony Chang <ac.chang@outlook.com>
[ROCm/composable_kernel commit: 6d8614ee50]
Composable Kernel
Methodology
Composable Kernel (CK) library aims to provide a programming model for writing performance critical kernels for machine learning workloads across multiple architectures including GPUs, CPUs, etc, through general purpose kernel languages, like HIP C++.
CK utilizes two concepts to achieve performance portability and code maintainability:
- A tile-based programming model
- Algorithm complexity reduction for complex ML operators, using innovative technique we call "Tensor Coordinate Transformation".
Code Structure
Current CK library are structured into 4 layers:
- "Templated Tile Operators" layer
- "Templated Kernel and Invoker" layer
- "Instantiated Kernel and Invoker" layer
- "Client API" layer
Contributors
The list of developers and contributors is here: Contributors
Citation
If you use CK, please use following citations:
- CK paper will be freely available on arXiv soon: Realizing Tensor Operators Using Coordinate Transformations and Tile Based Programming
- CITATION.cff
License
CK is released under the MIT license. License File
Build CK
Build docker image
DOCKER_BUILDKIT=1 docker build -t ck:latest -f Dockerfile .
Launch docker
docker run \
-it \
--privileged \
--group-add sudo \
-w /root/workspace \
-v ${PATH_TO_LOCAL_WORKSPACE}:/root/workspace \
ck:latest \
/bin/bash
Build CK
mkdir build && cd build
# Need to specify target ID, example below is for gfx908 and gfx90a
cmake \
-D CMAKE_PREFIX_PATH=/opt/rocm \
-D CMAKE_CXX_COMPILER=/opt/rocm/bin/hipcc \
-D CMAKE_CXX_FLAGS="-O3" \
-D CMAKE_BUILD_TYPE=Release \
-D GPU_TARGETS="gfx908;gfx90a" \
..
Build examples and tests
make -j examples tests
make test
Instructions for running each individual examples are under example
Build ckProfiler
make -j ckProfiler
Instructions for running ckProfiler are under profiler
Install CK
make install
Using CK as pre-built kernel library
Instructions for using CK as a pre-built kernel library are under client_example
Caveat
Kernel Timing and Verification
CK's own kernel timer will warn up kernel once, and then run it multiple times to get average kernel time. For some kernels that use atomic add, this will cause output buffer to be accumulated multiple times, causing verification failure. To work around it, do not use CK's own timer and do verification at the same time. CK's own timer and verification in each example and ckProfiler can be enabled or disabled from command line.

