Switch to standard ROCm packaging (#301)

* Switch to standard ROCm packaging

* Revert .gitignore changes

* install new rocm-cmake version

* update readme
Co-authored-by: illsilin <Illia.Silin@amd.com>
Co-authored-by: Chao Liu <chao.liu2@amd.com>

.gitignore

@@ -45,4 +45,4 @@ build*
 *~
 
 # GDB temporary files
-.gdb_history
+.gdb_history
\ No newline at end of file

CMakeLists.txt

@@ -7,7 +7,8 @@ list(APPEND CMAKE_MODULE_PATH "${PROJECT_SOURCE_DIR}/cmake")
 
 enable_testing()
 
-find_package(ROCM REQUIRED PATHS /opt/rocm)
+set(ROCM_SYMLINK_LIBS OFF)
+find_package(ROCM 0.8 REQUIRED PATHS /opt/rocm)
 
 include(ROCMInstallTargets)
 include(ROCMPackageConfigHelpers)
@@ -16,7 +17,7 @@ include(ROCMInstallSymlinks)
 include(ROCMCreatePackage)
 include(CheckCXXCompilerFlag)
 
-rocm_setup_version(VERSION 1.0.0)
+rocm_setup_version(VERSION 0.2.0)
 include(TargetFlags)
 list(APPEND CMAKE_PREFIX_PATH ${CMAKE_INSTALL_PREFIX} ${CMAKE_INSTALL_PREFIX}/llvm ${CMAKE_INSTALL_PREFIX}/hip /opt/rocm /opt/rocm/llvm /opt/rocm/hip)
 
@@ -70,7 +71,6 @@ if( DEFINED CK_OVERRIDE_HIP_VERSION_PATCH )
 endif()
 message(STATUS "Build with HIP ${HIP_VERSION}")
 
-
 rocm_create_package(
     NAME composablekernel
     DESCRIPTION "High Performance Composable Kernel for AMD GPUs"
@@ -238,6 +238,11 @@ message("CMAKE_CXX_FLAGS: ${CMAKE_CXX_FLAGS}")
 
 add_custom_target(check COMMAND ${CMAKE_CTEST_COMMAND} --output-on-failure -C ${CMAKE_CFG_INTDIR})
 
+rocm_package_setup_component(tests
+        LIBRARY_NAME composablekernel
+        PACKAGE_NAME tests # Prevent -static suffix on package name
+)
+
 add_subdirectory(library)
 add_subdirectory(example)
 add_subdirectory(test)
@@ -259,8 +264,19 @@ configure_package_config_file(${CMAKE_CURRENT_SOURCE_DIR}/Config.cmake.in
         NO_CHECK_REQUIRED_COMPONENTS_MACRO
 )
 
-install(FILES
+rocm_install(FILES
     "${CMAKE_CURRENT_BINARY_DIR}/composable_kernelConfig.cmake"
     "${CMAKE_CURRENT_BINARY_DIR}/composable_kernelConfigVersion.cmake"
     DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/composable_kernel
 )
+
+set(CPACK_RESOURCE_FILE_LICENSE "${CMAKE_CURRENT_SOURCE_DIR}/LICENSE")
+set(CPACK_RPM_PACKAGE_LICENSE "MIT")
+
+rocm_create_package(
+    NAME composablekernel
+    DESCRIPTION "High Performance Composable Kernel for AMD GPUs"
+    MAINTAINER "MIOpen Kernels Dev Team <dl.MIOpen@amd.com>"
+    LDCONFIG
+    HEADER_ONLY
+)

Dockerfile

@@ -88,3 +88,8 @@ ADD rbuild.ini /rbuild.ini
 ADD dev-requirements.txt dev-requirements.txt
 RUN rbuild prepare -s develop -d $PREFIX
 RUN groupadd -f render
+
+# Install the new rocm-cmake version
+RUN git clone -b master https://github.com/RadeonOpenCompute/rocm-cmake.git  && \
+  cd rocm-cmake && mkdir build && cd build && \
+  cmake  .. && cmake --build . && cmake --build . --target install

README.md

@@ -10,6 +10,9 @@ rocm/tensorflow:rocm5.1-tf2.6-dev              \
 /bin/bash
 ```
 
+# Install the new rocm-cmake version
+https://github.com/RadeonOpenCompute/rocm-cmake
+
 ## Build
 ```bash
 mkdir build && cd build
@@ -34,7 +37,7 @@ Instructions for running each individual examples are under ```example/```
 
 ## Tests
 ```bash
- make -j tests
+ make -j examples tests
  make test
 ```
 

cmake/googletest.cmake

@@ -8,7 +8,7 @@ endif()
 
 message(STATUS "Fetching GoogleTest")
 
-list(APPEND GTEST_CMAKE_CXX_FLAGS 
+list(APPEND GTEST_CMAKE_CXX_FLAGS
      -Wno-undef
      -Wno-reserved-identifier
      -Wno-global-constructors
@@ -31,7 +31,11 @@ FetchContent_Declare(
 
 # Will be necessary for windows build
 # set(gtest_force_shared_crt ON CACHE BOOL "" FORCE)
-FetchContent_MakeAvailable(googletest)
+FetchContent_GetProperties(googletest)
+if(NOT googletest_POPULATED)
+  FetchContent_Populate(googletest)
+  add_subdirectory(${googletest_SOURCE_DIR} ${googletest_BINARY_DIR} EXCLUDE_FROM_ALL)
+endif()
 
 target_compile_options(gtest PRIVATE ${GTEST_CMAKE_CXX_FLAGS})
 target_compile_options(gtest_main PRIVATE ${GTEST_CMAKE_CXX_FLAGS})

library/src/host_tensor/CMakeLists.txt

@@ -11,22 +11,20 @@ target_compile_features(host_tensor PUBLIC)
 set_target_properties(host_tensor PROPERTIES POSITION_INDEPENDENT_CODE ON)
 target_include_directories(host_tensor SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
 
-target_include_directories(host_tensor PUBLIC 
+target_include_directories(host_tensor PUBLIC
     "$<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}/ck>"
     "$<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}/ck/utility>"
     "$<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}/ck/library/host_tensor>"
 )
 
-install(TARGETS host_tensor 
-        EXPORT host_tensorTargets
-        LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
-        ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
-        RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR}
-        INCLUDES DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
+rocm_install(
+    TARGETS host_tensor
+    EXPORT host_tensorTargets
 )
 
-install(EXPORT host_tensorTargets
-    FILE composable_kernelhost_tensorTargets.cmake 
+rocm_install(
+    EXPORT host_tensorTargets
+    FILE composable_kernelhost_tensorTargets.cmake
     NAMESPACE composable_kernel::
     DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/composable_kernel
 )

library/src/tensor_operation_instance/gpu/CMakeLists.txt

@@ -75,21 +75,17 @@ target_include_directories(device_operations PUBLIC
 #once new arches are enabled make this an option on the main cmake file
 # and pass down here to be exported
 
-target_compile_options(device_operations PRIVATE 
+target_compile_options(device_operations PRIVATE
     --offload-arch=gfx908
     --offload-arch=gfx90a
 )
 
 # install(TARGETS device_operations LIBRARY DESTINATION lib)
-install(TARGETS device_operations
-        EXPORT device_operationsTargets
-        LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
-        ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
-        RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR}
-        INCLUDES DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
-)
-install(DIRECTORY ${DEV_OPS_INC_DIRS} DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/ck)
-install(EXPORT device_operationsTargets
+rocm_install(TARGETS device_operations
+        EXPORT device_operationsTargets)
+
+rocm_install(DIRECTORY ${DEV_OPS_INC_DIRS} DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/ck)
+rocm_install(EXPORT device_operationsTargets
         FILE composable_kerneldevice_operationsTargets.cmake
         NAMESPACE composable_kernel::
         DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/composable_kernel

library/src/tensor_operation_instance/gpu/conv2d_bwd_weight/CMakeLists.txt

@@ -3,9 +3,9 @@ set(DEVICE_CONV2D_BWD_WEIGHT_INSTANCE_SOURCE
    device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f16_instance.cpp;
    device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f32_instance.cpp;
 )
-add_library(device_conv2d_bwd_weight_instance OBJECT ${DEVICE_CONV2D_BWD_WEIGHT_INSTANCE_SOURCE}) 
+add_library(device_conv2d_bwd_weight_instance OBJECT ${DEVICE_CONV2D_BWD_WEIGHT_INSTANCE_SOURCE})
 target_compile_features(device_conv2d_bwd_weight_instance PUBLIC)
 set_target_properties(device_conv2d_bwd_weight_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
-install(TARGETS device_conv2d_bwd_weight_instance LIBRARY DESTINATION lib) 
+rocm_install(TARGETS device_conv2d_bwd_weight_instance)
 
 clang_tidy_check(device_conv2d_bwd_weight_instance)

library/src/tensor_operation_instance/gpu/convnd_bwd_data/CMakeLists.txt

 # device_convnd_bwd_data_instance
-set(DEVICE_CONVND_BWD_DATA_INSTANCE_SOURCE 
+set(DEVICE_CONVND_BWD_DATA_INSTANCE_SOURCE
    device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f16_instance.cpp;
    device_conv1d_bwd_data_xdl_nwc_kxc_nwk_f32_instance.cpp;
    device_conv1d_bwd_data_xdl_nwc_kxc_nwk_bf16_instance.cpp;
@@ -12,11 +12,11 @@ set(DEVICE_CONVND_BWD_DATA_INSTANCE_SOURCE
    device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_f32_instance.cpp;
    device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_bf16_instance.cpp;
    device_conv3d_bwd_data_xdl_ndhwc_kzyxc_ndhwk_int8_instance.cpp;
-) 
+)
 
 add_library(device_convnd_bwd_data_instance OBJECT ${DEVICE_CONVND_BWD_DATA_INSTANCE_SOURCE})
 target_compile_features(device_convnd_bwd_data_instance PUBLIC)
 set_target_properties(device_convnd_bwd_data_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
-install(TARGETS device_convnd_bwd_data_instance LIBRARY DESTINATION lib) 
+rocm_install(TARGETS device_convnd_bwd_data_instance)
 
 clang_tidy_check(device_convnd_bwd_data_instance)

library/src/tensor_operation_instance/gpu/gemm_bias_add_reduce/CMakeLists.txt

@@ -6,5 +6,5 @@ set(DEVICE_GEMM_REDUCE_INSTANCE_SOURCE
 )
 
 add_instance_library(device_gemm_bias_add_reduce_instance ${DEVICE_GEMM_REDUCE_INSTANCE_SOURCE})
-install(TARGETS device_gemm_bias_add_reduce_instance LIBRARY DESTINATION lib)
+rocm_install(TARGETS device_gemm_bias_add_reduce_instance)
 clang_tidy_check(device_gemm_bias_add_reduce_instance)

library/src/tensor_operation_instance/gpu/gemm_reduce/CMakeLists.txt

@@ -6,5 +6,5 @@ set(DEVICE_GEMM_REDUCE_INSTANCE_SOURCE
 )
 
 add_instance_library(device_gemm_reduce_instance ${DEVICE_GEMM_REDUCE_INSTANCE_SOURCE})
-install(TARGETS device_gemm_reduce_instance LIBRARY DESTINATION lib)
+rocm_install(TARGETS device_gemm_reduce_instance)
 clang_tidy_check(device_gemm_reduce_instance)

library/src/tensor_operation_instance/gpu/grouped_gemm/CMakeLists.txt

@@ -6,10 +6,10 @@ set(DEVICE_GROUPED_GEMM_INSTANCE_SOURCE
    device_grouped_gemm_xdl_f16_f16_f16_km_nk_mn_instance.cpp;
 )
 
-add_library(device_grouped_gemm_instance OBJECT ${DEVICE_GROUPED_GEMM_INSTANCE_SOURCE}) 
+add_library(device_grouped_gemm_instance OBJECT ${DEVICE_GROUPED_GEMM_INSTANCE_SOURCE})
 
 target_compile_features(device_grouped_gemm_instance PUBLIC)
 set_target_properties(device_grouped_gemm_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
-install(TARGETS device_grouped_gemm_instance LIBRARY DESTINATION lib)
+rocm_install(TARGETS device_grouped_gemm_instance)
 
 clang_tidy_check(device_grouped_gemm_instance)

test/CMakeLists.txt

@@ -13,6 +13,7 @@ function(add_test_executable TEST_NAME)
     add_test(NAME ${TEST_NAME} COMMAND $<TARGET_FILE:${TEST_NAME}> )
     add_dependencies(tests ${TEST_NAME})
     add_dependencies(check ${TEST_NAME})
+    rocm_install(TARGETS ${TEST_NAME} COMPONENT tests)
 endfunction(add_test_executable TEST_NAME)
 
 include(GoogleTest)
@@ -26,6 +27,7 @@ function(add_gtest_executable TEST_NAME)
     target_compile_options(${TEST_NAME} PRIVATE -Wno-global-constructors -Wno-undef)
     target_link_libraries(${TEST_NAME} PRIVATE gtest_main)
     gtest_discover_tests(${TEST_NAME})
+    rocm_install(TARGETS ${TEST_NAME} COMPONENT tests)
 endfunction(add_gtest_executable TEST_NAME)