mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-20 04:49:54 +00:00
87522d2ac9
* adding batched_gemm_and_reduction
* batched_gemm_reduce works with bactch_count=1
* fix a bug in grid_size; batched_gemm_reduce works for batch_count > 1
* adding profiler for batched_gemm_fp16
* fixed a bug in declaration of d1 and d0; both example and profiler work
* clang-format
* cleanup
* batched_gemm_reduce: add test
* minor change
* fixed some typo in function names
[ROCm/composable_kernel commit: 34c661e71c]
59 lines
2.8 KiB
CMake
59 lines
2.8 KiB
CMake
include_directories(BEFORE
|
|
${PROJECT_SOURCE_DIR}/include/ck
|
|
${PROJECT_SOURCE_DIR}/include/ck/utility
|
|
${PROJECT_SOURCE_DIR}/include/ck/tensor_description
|
|
${PROJECT_SOURCE_DIR}/include/ck/tensor
|
|
${PROJECT_SOURCE_DIR}/include/ck/problem_transform
|
|
${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/device
|
|
${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/grid
|
|
${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/block
|
|
${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/warp
|
|
${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/thread
|
|
${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/element
|
|
${PROJECT_SOURCE_DIR}/library/include/ck/library/host_tensor
|
|
${PROJECT_SOURCE_DIR}/library/include/ck/library/tensor_operation_instance
|
|
${PROJECT_SOURCE_DIR}/library/include/ck/library/tensor_operation_instance/gpu/reduce
|
|
${PROJECT_SOURCE_DIR}/library/include/ck/library/reference_tensor_operation/cpu
|
|
${PROJECT_SOURCE_DIR}/library/include/ck/library/reference_tensor_operation/gpu
|
|
${PROJECT_SOURCE_DIR}/profiler/include
|
|
${PROJECT_SOURCE_DIR}/external/include/half
|
|
)
|
|
|
|
# ck_profiler
|
|
set(PROFILER_SOURCE
|
|
src/profiler.cpp
|
|
src/profile_gemm.cpp
|
|
src/profile_gemm_bias_2d.cpp
|
|
src/profile_gemm_bias_relu.cpp
|
|
src/profile_gemm_bias_relu_add.cpp
|
|
src/profile_gemm_reduce.cpp
|
|
src/profile_batched_gemm.cpp
|
|
src/profile_conv_fwd.cpp
|
|
src/profile_conv_fwd_bias_relu.cpp
|
|
src/profile_conv_fwd_bias_relu_add.cpp
|
|
src/profile_conv_fwd_bias_relu_atomic_add.cpp
|
|
src/profile_convnd_bwd_data.cpp
|
|
src/profile_reduce.cpp
|
|
src/profile_grouped_gemm.cpp
|
|
src/profile_batched_gemm_reduce.cpp
|
|
)
|
|
|
|
add_executable(ckProfiler ${PROFILER_SOURCE})
|
|
|
|
target_link_libraries(ckProfiler PRIVATE host_tensor)
|
|
target_link_libraries(ckProfiler PRIVATE device_gemm_reduce_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_gemm_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_gemm_bias2d_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_gemm_bias_relu_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_gemm_bias_relu_add_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_batched_gemm_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_bias_relu_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_bias_relu_add_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_bias_relu_atomic_add_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_convnd_bwd_data_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_reduce_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_reduce_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_grouped_gemm_instance)
|
|
target_link_libraries(ckProfiler PRIVATE device_batched_gemm_reduce_instance)
|