* dump lds content in appropriate precision type
* add squared add reduction op; allows sq sum
* initial stub from regular gemm impl
* layernorm example code & host verification
* initial layernorm implementation
* tidy up
* make C0 precision type consistent with C
* clang-tidy and additional comments
* tighten up example code
* account for extra flops/bytes from normalization
* clang-format
* c0 bias/beta/gamma now have its own precision type
* AccElemOp for gemm outputs prior to feeding to layernorm
* update workgroup mapping
* rename kernel template param to reflect its dual use
* use LDS mem pool for reduction workspace
* change cshuffle precision type to f16; clean up
* clang-format
* correct naming
* explicit cast
* fully implemented gemm + bias + activation + add + norm
* activation in correct order
* reflect reduction API's recent change
* amend
* clean up; add comment
* keep up with recent changes in reduction API
* format
* resolve merge conflicts
Co-authored-by: Chao Liu <chao.liu2@amd.com>
* use 'sweep once' softmax kernel where applicable
* threadwise copy's dst buffer can specify invalid element value
* add int8 in/out float compute softmax support
give a bit of leeway for int absolute tolerance as there's a single data point of all test cases showing off-by-1 error
* format
* softmax inherits DeviceNormalization
* softmax profiler stub
* tighten up reference softmax interface
* example prints tensor dimension
* add fp32 to softmax profiler
* rename header
* hook with ckProfiler
* format
* resolve merge conflict
* resolve merge conflicts
* update normalization profiler help string
* resolve conflict
* typo
* remove residual
* softmax profiler: address feedback
* test for mixed precision input/output
* fully qualify ck::math::isnan
* add comment for device normalization interface
* revise wording
* constness for alpha/beta scaler pointer
* Extract base class for elementwise
* Refactor interface of DeviceGemmReduce. Do not use tuple in interface
* [What] Rename d into reduce in gemm + reduction related code
[Why] Prepare to add d term for add
* Unify base class of gemm + reduce and gemm + bias + add + reduce
* 1. Rename gemm_bias_add_reduce for external api
2. Refine cmake
* Add normalize device operation
* [What] Reorder the argument
[Why] Because d0 is also the input of c.
* Add type string
* Add example of gemm_bias_add_layernorm via external api
* Refactor example code
* clang-format
* Fix compile error
* clang-format
* Add external api for gemm_add_add_layernorm and normalize
* Add client example
* clang-format
* initial stub for standalone softmax
* start device_softmax_mk_to_mk as a wrapper to device_reduce_mk_to_m
* host softmax validates
* compiles; to implement beta scaling
* use NaN trick to efficiently ignore OOB values during sum of exponentials
* freeload device_reduce's utility functions
* clean up interface
* adding prior value (beta scaling)
* remove restriction related to perf considerations
* apply clang-format
* clean; disable diagnostics
* resolve conflicts
* add exp wrapper
* honor HostTensorDesc interface; allow implicit cast from different vector<T> type
* test softmax for fp16/fp32
* update readme
* amend commit NaN trick
* remove redundant param added during development
* format
* replace ScalarDataType with AccDataType
* separate out test programs by precision type
* move softmax sample code to its own folder
* format
* keep up with recent changes in reduction API
* remove extra header
* Remove template from Reducton operation classes and add template to their operator() and GetIdentityValue() interfaces
* Change to unary elementwise operators and the reduce_unary_operator (class for mapping) and dependent variations in all host layers
* Remove the data type template parameter from reduce_binary_operator (class for mapping) and dependent variations in host layers
* Add InMemoryDataOperatonSupportedOnDataType to check the matching between data type and InMemoryDataOperation
* Use struct-scope operator template instantiation for binary and unary element-wise operations
* Change a few more elementwise operations to use template for operator()
* Tiny correction in Normalize operator
* Add static_assert to check the data type appliability for some reduction accumulator and element-wise operatons
* Correction in some examples with regard to using ReduceAccDataType
* Use static_assert for UnaryDivide
* Update to merged codes to use Element-wise operations and Reduction Accumulator operations correctly
* Tiny fix with regard to SetWorkSpacePointer()
* Copy "gemm reduce" to "gemm bias add reduce"
* Implement gemm bias add reduction
* Fix compiler error due to merge from develop
* Add tensor operation for gemm + bias + add + reduce
* Add gemm_bais_add_reduce to ckProfiler
* Add c1 functor
* Refine type
* Use reduceAccDataType instead of explicitly float
* Change to use check_err()
* Do relu in float32 instead of bhalf_t. Because bhalf_t is unsigned
* Refactor relu. using type_trait instead of overloading
* Rename DxsReduceAccElementwiseOperation to DxsReduceAccElementwiseOperation
* Fix denominator
* Refine nameing
* Fix denominator in host
* Remove useless include header
* Use AccDataType
* Fix static_cast order
* Refine type
* [What] Remove tuple type in the base class
[Why] External api depend on base class. if base class has relationship with type, we will need many class for different type
* add GetWorkSpaceSize to base arg and make an example on convnd_bwd_weight
* add bwd weight for bf16: init
* remove redundant compute
* use datatype and split k to check whether a workspace is used
* remove unused computation for work space size
* add some code for bfp16
* add device/grid unary op
* add unary type convert to bwd-weight example
* support bf16 splitk kernel for convnd bwd weight
* 1. remove comments. 2. add checkvalidity. 3. add gridsize computation
* add workspace size check
* fix format
* change function name
* Use the unified naming for math functions on host and HIP kernel
* Corresponding change/simplification in reduction host/profiler/examples due to unified math functions renaming
* Renaming GetReductionZeroVal() to GetIdentityValue()
* Tiny renaming in profile_reduce_impl.hpp
* More renaming in profile_reduce_impl.hpp
* Replace zeroVal by identiyVal
* Remove ck_ prefix in the naming of ck::math provided functions
* Implement reduction meand and reduction square mean
* Refine file name
* Add reduce mean and square mean
* Fix parameter name
* Add normalize device op (not implement invoker::run())
* Remove epislon
* Refine deviceop
* Add 5ary elementwise for normalization
* Add layernorm example
* layerNorm verication
* Fix compiler error due to merge from develop
* Fix typo
* Fix compile error
* Refine naming
* [What] Suport non pointer for invoker and argument
[Why] Snyc coding style with gemm
* Refine folder name
* Refine class name
* Evaluate perf of the kernel
* Fix compile error
* [What] Refine perf evaluation in example of gemm + reduction
[Why] evaluation of gemm + reduction may cause verification fail. Because evaluation will not initial global memory
* clang-format
* debugging conv
* fix oversight where ctile map is constructed before initializing c desc
* example program should returns error code
* clean up
* changed Block2CTileMap in conv2d and convnd
* clean up
* clean up
* cleanup
Co-authored-by: Anthony Chang <ac.chang@outlook.com>
* Support different length of ScalarPerVector
* Add example of broadcast on fastest axis
* Typo
* Refine fastest example
* Add dimension check
* Modify fastest broadcast example to 3d
* Enforce users give scalarPerVector explicitely
* 1. Add CscalarPerVedctor
2. Not only broadcast on fastest need to set scalarPerVector to 1
* Rename var
* Move IsScalarPerVectorValid() inside IsSupportedArgument()
* Separate GridDesc_M0 into A, B and C
* rename var
* Rename var of length
Co-authored-by: rocking <chunylai@amd.com>
* start adding navi21 GEMM
* navi_gemm_km_kn_mn_fp32 compiles and passes one test.
* rename variables and functions in gridwise_gemm_dlops_v1r3
* add other 3 layouts; format instance
* adding more tuning parameters
add tuning parameters for other 3 layouts
* add gemm_dlops_f16
* tmp
* add dependence of DeviceGemm::IsSupportedArg() on arch
* minor changes
* minor changes
* minor changes
* minor changes
* minor changes
* minor changes
* minor changes
* push gemm_dlops into profiler
* minor changes
* if using xdl or dlops is moved into profiler_gemm_impl
* minor changes
* minor changes
* remove is_xdl from profile_gemm_impl
* make IsSupportedArg dependent on arch for other device_gemm
* minor changes
* minor changes
* fix a bug in f_generate_tensor_value
* add 64x64x64 for gemm_dlops_int8
* add 64x64x64 for gemm_dlops_int8
* comment out 3 layouts in gemm_dlops_int8; add 32x32x32 for gemm_dlops_int8; init A values to 1
* fix
* start fixing tuning parameters
* monir
* minor changes
* minor changes
* minor changes
* fixing
* adding example
* adding example
* adding example
* add gemm fp32 example
* clean up
* use 128x128x16 as MNK tile in navi21 gemm example
* bug fix
* fix test
* use new block c tile
* clean
* fix build
Co-authored-by: Chao Liu <chao.liu2@amd.com>
Co-authored-by: shaojiewang <wsjmessi@163.com>
* Tiny fix in dynamic_buffer.hpp to support vectorized AtomicAdd for double type
* Update to host layer and host reduction
* Merge and remove reduction kernels
* Merge and remove reduction device interfaces and update pooling device interface
* Merge and remove useless reduction device instances
* Update to reduction profiler and reduction ctests
* Update to reduction and pooling examples and add one reduction example
* Change to reduction examples to let them testable by ctest
* Add explicit pass checking for reduction and pooling examples
* Explicit assignment of tensor shapes in example reduce_blockwise_two_call
* Use atomic_add to repace atomicAdd and add atomic_add for double type
* Add reduce ctest support for double data type
* Replace to_int_vector() by using c++ std::vector::assign()
* Keep DeviceReduceThreadWise separated from DeviceReduceBlockWise
* Merge DeviceReduceBlockWise and DeviceReduceMultiBlockAtomicAdd into DeviceReduceMultiBlock
* Add GetAtomicOperationZeroValue() support for AtomicMax
* Tiny change to reduce example README.md
* Fix some tiny issues due to branch merging
* Revoke previous change in dynamic_buffer.hpp and add atomic_add for double2_t
* Add reduce multiblock_atomic_add instances for fp64 to verify vectorized atomic_add on fp64
* Renaming
* Clean the header includings in device_reduce instances header files
* add some instance to develop
* avoid bank conflicts for wrw for all instance
* add small K1 test
* delete some unused instance
* binding gemm k1 to conv n
* try using half_4 to do ds_read
* reset buffer load oob and ds memcpy to default option
* remove useless instances
* remove redandunt space
* remove printf code
* clang-format-10 change
* use fastest config
* fix clang format for the other files
* remove gemmk0 pad for output
* add gemmk padding macro
* add bank length computation
* add template to distinguish the instance that need lds padding for wrw
* use rocm5.1 as docker
* use integer value for GEMM test
* add Right padding macro
* add 2 test asm code
* using 256x256x32 tile size
* 1. move dedicated transform into gridwisegemm's head file. 2. make lds tensor params a struct templete. 3. remove useless code
* using small vec
* 256*128 kernel size for example
* remove asm files
* use a new gridwise gemm header for bwd-weight
* revert gridwise gemm v2r4r2
* change foramt
* reset gridwise gemm v2r4r2
* remove unused code
* revert instance file
* revert example instance
* format file
* remove macros
* resolve compile error
* rename wrw kernel invoker
* use gridwisegemm pipeline struct instead of implement run fucntion in the same header
Co-authored-by: Chao Liu <chao.liu2@amd.com>
* add some instance to develop
* avoid bank conflicts for wrw for all instance
* add small K1 test
* delete some unused instance
* reset buffer load oob and ds memcpy to default option
* remove useless instances
* remove redandunt space
* remove printf code
* clang-format-10 change
* fix clang format for the other files
* add bank length computation
* add template to distinguish the instance that need lds padding for wrw
* use rocm5.1 as docker
* use integer value for GEMM test
* 1. move dedicated transform into gridwisegemm's head file. 2. make lds tensor params a struct templete. 3. remove useless code
* use a new gridwise gemm header for bwd-weight
* revert gridwise gemm v2r4r2
* change foramt
* rename kernel invoker
Co-authored-by: Chao Liu <chao.liu2@amd.com>
* [What] Rename the example
[Why] Prepare to add unary reduction
* Add global oparation to the parameter
* Add atomicmax
* Fix compile error
* Support atomicMax (hip library)
* Rename the reduction example
* Fix target name
* use p_d1_grid as the indicator directly
* Prevent performance issue. Let passthrough handle it.
* Implement the function template the specialize the float2
* No need to separate into two lines
* Remove empty line
* add comment
* Fix compile error due to merge from develop
* make the implementation of atomic_max / atomic_add explicit for each datatype
* Refine typo
* For future CI test
* Fix compiler error in ckProfiler
* Merge commit 'de2769e3a6695b38a20529261273ddc5cdaab2fe'
* simply use remove_pointer
* Rename type and var
* Refine example
* Modify reducemax example
* Fix bug in reduction
* Change initialize range
* Implement F64 version of atomicMax
* Move reduction code together
* Add buffer atomic_max
* Fix coding style by clang-format
* Integrate new api of DeviceGemmReduce_Xdl_CShuffle
* Integrate Batch gemm reduction
* Fix example
* fix example
* clean up
* Fix batch gemm tensor operation
* Fix coding style
* Fix template augument
* Fix clang format
* Keep flexible of different stride for each D tensor
* Fix compile error for ckProfiler
* Fix typo
* [What] Fix naming
[Why] Prepare to add out elementop
* Add DoutElementOp
Co-authored-by: Chao Liu <chao.liu2@amd.com>
Co-authored-by: rocking <chunylai@amd.com>
* Add elementwise operation kernel and example
* Add comment
* Add template argument of dim . Prepare to support multiple dimension
* Rename example
* Support 1 dimension
* Add static assert
* Add comment
* Extract pad
* Remove redundant argument
* Support any dimension for elementwise operation
* Remove line
* Let it be the multiple number of CU
* Move thread per block to the parameter of constructor
* rename threadPerBlock with blockSize
* Support double
* rename kernel function name
* remove redundant include header
* Refine type
* Need to the final dimension
* Refine variable name
* Refine type
* Use index_t instead of int in API
Co-authored-by: rocking <chunylai@amd.com>
* manual control of MAC cluster for improved 2-wave performance
ensure setprio's order; ensure inner loop size >= local read size
synchronize when single mac cluster
* format
* use value field from ck::integral_constant
* roll out inter-wave loop scheduler to c-shuffle gemm variants
will gradually roll out to other applicable device ops when occasional reg spill is resolved
* additional comments
* format
* fix mismatch between inter-wave pipeline and interwave blockwise gemm
* address review feedback
* amend
* [Experimental] Change to gemm+reduce and batched-gemm+reduce
* Use threadwise-reduce function to improve the gridwise_gemm_reduce_xdl_cshuffle kernel
* Tiny fix in device_batched_gemm_xdl.hpp
* clang-format library/src/utility/conv_fwd_util.cpp
* compile ck for all targets
* update the target criteria
* change the target condition
* fixed some typos
* fixed missed file
* revert changes in README
* revert device_conv3d_fwd_xdl_...
* update device_conv3d_fwd_xdl_...
* update device_batched_gemm_reduce...
* test the unused arguments fix
* test the warning suppression
* try suppress warnings in device_batched_gemm_reduce_xdl...
* fix the last warnings
* replace UNUSED with std::ignore
* fix a typo
* replaced std::ignore with ignore
* add igonre header to common_header
* refactor atomicAdd
Co-authored-by: Chao Liu <chao.liu2@amd.com>
* Add ThreadwiseReduction functor as per-thread reduction api
* Using ThreadwiseReduce api and some change in using PartitionedBlockwiseReduction api to simply the kernels
* Add comments and remove useless declarations in the kernels
* Tiny updates
* init of grouped_gemm
* 2 gemm test
* perf test
* clean
* wrap desc into a struct
* test cast static_arr to pointer
* add ptr to GemmDesc
* add grouped gemm profiler
* fixed mem issue with unique_ptr
* clean
* clean
* finished ckprofiler
* Update README.md
* readme
* fixed readme
* add example
* improve code
* fixed comments: reserve, seperate ptr and gemm_shapes
* merge group and non-group
* fixed comments: replace push_back with emplace_back to avoid copy constructor
* fixed comments: unified blk2ctile; add test
* ci fix
* fixed ci
* fixed ci
* fixed ci
* Use thread cluster descriptor and explicit M_K 2d descriptor to simply Blockwise Reduction
* Change by replacing ReduceDims by NumReduceDims as Device Reduce interface template parameter
* Rename the folder name for the pool2d and reduce examples
* Update to reduction test scripts
* Add Readme for pool2d_fwd and reduce_blockwise examples
* Add support for int8_t reduction (ADD/AVG, MIN/MAX/AMAX)
* Tiny fix in reduce profiler and tiny update in reduce testing scripts
* Tiny fix in testing script profile_reduce_no_index.sh
* Tiny fix in testing script profile_reduce_no_index.sh
* Add support for bfp16 reduction (using bhalf_t = ushort)
* Tiny fix in amd_buffer_addressing.hpp
* Tiny change in script/profile_reduce_with_index.sh
* Use AccDataType for Beta value and use element_wise::PassThrough
* Use type_convert for type converting in host layer reduction
* Renaming and refining in Reduction profiler/device layer/examples
* Renaming and refining in Reduction profiler/device layer/examples
* Renaming all NumReduceDims to NumReduceDim
* Fix the leaked type_convert in ThreadwiseTensorSliceTransfer_v2
* Update to testing scripts to add bf16 support
* added more static_assert
* Remove buggy tunable configurations defined in device_reduce_instance_xxx.hpp
* Add static_assert to give compile-time warning for incorrect thread slice-size/vector-size configurations
* minor change
* Refine and fix (in GetWorkspaceSizeInBytes of MultiBlockPartialReduce) to make int8 completely pass
* Tiny renaming in gridwise_2d_reduction_multiblock_partial_reduce.hpp
* Tiny fix in script/profile_reduce_no_index.sh
* Refine in DeviceReduce layer with regard to using NumInvariantDim/NumReduceDim or InvariantDims/ReduceDims
* Generic renaming in host reduction and DeviceReduce layer
* Add support for 4-d all dimension reduction in the profiler and add_device_reduce_xxx instances
* Use multi-thread and simplification for host Reduction implementation
* Add ctest for reduction
* Update to clarify the using of data init method in produce_reduce/example_reduce/test_reduce/
* Update to the reduce CTest executables to enable default testing behavior when no command argument
* Renaming
Co-authored-by: Jianfeng yan <jfyan008@gmail.com>
* Use thread cluster descriptor and explicit M_K 2d descriptor to simply Blockwise Reduction
* Change by replacing ReduceDims by NumReduceDims as Device Reduce interface template parameter
* Rename the folder name for the pool2d and reduce examples
* Update to reduction test scripts
* Add Readme for pool2d_fwd and reduce_blockwise examples
* Tiny fix in reduce profiler and tiny update in reduce testing scripts
* Tiny fix in testing script profile_reduce_no_index.sh
* Tiny change in script/profile_reduce_with_index.sh
* Renaming and refining in Reduction profiler/device layer/examples
* Renaming and refining in Reduction profiler/device layer/examples
* Renaming all NumReduceDims to NumReduceDim
