mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-13 09:45:56 +00:00
34c661e71c
* 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
46 lines
1.9 KiB
CMake
46 lines
1.9 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/reference_tensor_operation/cpu
|
|
${PROJECT_SOURCE_DIR}/library/include/ck/library/reference_tensor_operation/gpu
|
|
${PROJECT_SOURCE_DIR}/external/include/half
|
|
)
|
|
|
|
add_custom_target(examples)
|
|
|
|
function(add_example_executable EXAMPLE_NAME)
|
|
message("adding example ${EXAMPLE_NAME}")
|
|
add_executable(${EXAMPLE_NAME} ${ARGN})
|
|
target_link_libraries(${EXAMPLE_NAME} PRIVATE host_tensor)
|
|
add_dependencies(examples ${EXAMPLE_NAME})
|
|
endfunction(add_example_executable EXAMPLE_NAME)
|
|
|
|
add_subdirectory(01_gemm)
|
|
add_subdirectory(02_gemm_alpha_beta)
|
|
add_subdirectory(03_gemm_bias_relu)
|
|
add_subdirectory(04_gemm_bias_relu_add)
|
|
add_subdirectory(05_conv2d_fwd)
|
|
add_subdirectory(06_conv2d_fwd_bias_relu)
|
|
add_subdirectory(07_conv2d_fwd_bias_relu_add)
|
|
add_subdirectory(08_conv3d_fwd)
|
|
add_subdirectory(09_convnd_fwd)
|
|
add_subdirectory(10_conv2d_bwd_data)
|
|
add_subdirectory(11_conv2d_bwd_wgt)
|
|
add_subdirectory(12_reduce)
|
|
add_subdirectory(13_pool2d_fwd)
|
|
add_subdirectory(14_gemm_xdl_requant_relu_requant)
|
|
add_subdirectory(17_convnd_bwd_data_xdl)
|
|
add_subdirectory(15_grouped_gemm)
|
|
add_subdirectory(16_gemm_reduce)
|
|
add_subdirectory(18_batched_gemm_reduce)
|