mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-26 08:00:13 +00:00
a3c89131fa
* tuning para,
* testing on v100
* add fp16
* remove deprecated tensor descriptor
* sync with miopen
* update build script
Co-authored-by: Jing Zhang <jizhan@amd.com>
[ROCm/composable_kernel commit: 5c7cec1115]
31 lines
942 B
CMake
31 lines
942 B
CMake
set(TENSOR_SOURCE
|
|
src/host_tensor.cpp;
|
|
src/device.cpp;
|
|
)
|
|
|
|
add_library(host SHARED ${TENSOR_SOURCE})
|
|
target_compile_features(host PUBLIC)
|
|
set_target_properties(host PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
|
|
if(DEVICE_BACKEND STREQUAL "NVIDIA")
|
|
target_link_libraries(host nvToolsExt cudart)
|
|
endif()
|
|
|
|
install(TARGETS host LIBRARY DESTINATION lib)
|
|
|
|
|
|
if(DEVICE_BACKEND STREQUAL "AMD")
|
|
set(CONV_SOURCE src/conv_driver.cpp)
|
|
set(COL2IM_SOURCE src/col2im_driver.cpp)
|
|
set(CONV_BWD_DATA_SOURCE src/conv_bwd_data_driver.cpp)
|
|
elseif(DEVICE_BACKEND STREQUAL "NVIDIA")
|
|
set(CONV_SOURCE src/conv_driver.cu)
|
|
set(COL2IM_SOURCE src/col2im_driver.cu)
|
|
set(CONV_BWD_DATA_SOURCE src/conv_bwd_data_driver.cu)
|
|
endif()
|
|
|
|
add_executable(conv_driver ${CONV_SOURCE})
|
|
add_executable(conv_bwd_data_driver ${CONV_BWD_DATA_SOURCE})
|
|
target_link_libraries(conv_driver PRIVATE host)
|
|
target_link_libraries(conv_bwd_data_driver PRIVATE host)
|