feat(cmake/windows/cuda): upgrade windows llvm to

lastest 12.0.1, for fit cutlass upgrade

GitOrigin-RevId: 23b9d779dfbfc6dbbf31b63abdcc7a4aa6c654cd

scripts/cmake-build/BUILD_README.md

@@ -24,17 +24,16 @@
 	* install to default dir:  /c/Program\ Files/Git
 2: install visual studio 2019 Enterprise (Windows GUI)
 	* download install exe from https://visualstudio.microsoft.com
-	* choose "c++ develop" -> choose cmake/MSVC/clang/cmake/windows-sdk when install
+	* choose "c++ develop" -> choose cmake/MSVC/cmake/windows-sdk when install
 	* NOTICE: windows sdk version >=14.28.29910 do not compat with CUDA 10.1, please
 		choose version < 14.28.29910
 	* then install choosed components
-	* after install visual studio 2019 Enterprise, time to replace lld-link.exe
-		caused by visual studio 2019 lld-link.exe have crash issue
-		download office exe from https://releases.llvm.org/download.html
-		install to default: C:\Program Files\LLVM
-		cd "/c/Program Files (x86)/Microsoft Visual Studio/2019/Enterprise/VC/Tools/Llvm/bin"
-		cp /c/Program\ Files/LLVM/bin/lld-link.exe lld-link.exe
-3: install python3 (Windows GUI)
+3: install LLVM from https://releases.llvm.org/download.html (Windows GUI)
+    * llvm install by Visual Studio have some issue, eg, link crash on large project, please use official version
+    * download install exe from https://releases.llvm.org/download.html
+	* our ci use LLVM 12.0.1, if u install other version, please modify LLVM_PATH
+    * install 12.0.1 to /c/Program\ Files/LLVM_12_0_1
+4: install python3 (Windows GUI)
 	* download python 64-bit install exe (we support python3.5-python3.8 now)
 	     https://www.python.org/ftp/python/3.5.4/python-3.5.4-amd64.exe
 	     https://www.python.org/ftp/python/3.6.8/python-3.6.8-amd64.exe
@@ -52,21 +51,21 @@
 		python3.exe -m pip install --upgrade pip
 		python3.exe -m pip install -r imperative/python/requires.txt
 		python3.exe -m pip install -r imperative/python/requires-test.txt
-4: install cuda components (Windows GUI)
+5: install cuda components (Windows GUI)
 	* now we support cuda10.1+cudnn7.6+TensorRT6.0 on Windows
 	* install cuda10.1 to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1
 	* install cudnn7.6 to C:\Program Files\NVIDIA GPU Computing Toolkit\cudnn-10.1-windows10-x64-v7.6.5.32
 	* install TensorRT6.0 to C:\Program Files\NVIDIA GPU Computing Toolkit\TensorRT-6.0.1.5
-5: edit system env variables (Windows GUI)
+6: edit system env variables (Windows GUI)
 	* create new key: "VS_PATH", value: "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise"
+	* create new key: "LLVM_PATH", value: "C:\Program Files\LLVM_12_0_1"
 	* append "Path" env value
         C:\Program Files\Git\cmd
 		C:\Users\build\mge_whl_python_env\3.8.3
 		C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\bin
 		C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\libnvvp
 		C:\Program Files\NVIDIA GPU Computing Toolkit\cudnn-10.1-windows10-x64-v7.6.5.32\cuda\bin
-		C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Tools\Llvm\lib\clang\11.0.0\lib\windows
-		C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Tools\Llvm\x64\lib\clang\11.0.0\lib\windows
+		C:\Program Files\LLVM_12_0_1\lib\clang\12.0.1\lib\windows
 ```
 
 ### Linux host build

scripts/cmake-build/host_build.sh

@@ -161,7 +161,7 @@ function prepare_env_for_windows_build() {
     fi
     echo $VS_PATH
 
-    # only use cmake/clang-cl/Ninja install from Visual Studio, if not, may build failed
+    # only use cmake/Ninja install from Visual Studio, if not, may build failed
     # some user env may install cmake/clang-cl/Ninja at windows-git-bash env, so we put Visual Studio
     # path at the head of PATH, and check the valid
     echo "check cmake install..."
@@ -176,14 +176,22 @@ function prepare_env_for_windows_build() {
     fi
 
     echo "check clang-cl install..."
-    export PATH=$VS_PATH/VC/Tools/Llvm/bin/:$PATH
-    which clang-cl
+    # llvm install by Visual Studio have some issue, eg, link crash on large project, so we
+    # use official LLVM download from https://releases.llvm.org/download.html
+    if [[ -z ${LLVM_PATH} ]];then
+        echo "can not find LLVM_PATH env, pls export you LLVM install dir to LLVM_PATH"
+        echo "examle for export LLVM_12_0_1"
+        echo "export LLVM_PATH=/c/Program\ Files/LLVM_12_0_1"
+        exit -1
+    fi
+    echo ${LLVM_PATH}
+    export PATH=${LLVM_PATH}/bin/:$PATH
     clang_loc=`which clang-cl`
     if [[ $clang_loc =~ "Visual" ]]; then
-        echo "clang-cl valid ..."
-    else
-        echo "clang-cl Invalid: ..."
+        echo "clang-cl Invalid: we do not support use LLVM installed by Visual Studio"
         windows_env_err
+    else
+        echo "clang-cl valid ..."
     fi
 
     echo "check Ninja install..."

scripts/whl/windows/windows_build_whl.sh

@@ -16,8 +16,10 @@ function err_env() {
 function append_path_env_and_check() {
     echo  "export vs2019 install path"
     export VS_PATH=/c/Program\ Files\ \(x86\)/Microsoft\ Visual\ Studio/2019/Enterprise
+    echo  "export LLVM install path"
+    export LLVM_PATH=/c/Program\ Files/LLVM_12_0_1
     # for llvm-strip
-    export PATH=$VS_PATH/VC/Tools/Llvm/bin/:$PATH
+    export PATH=${LLVM_PATH}/bin/:$PATH
 }
 
 append_path_env_and_check

sdk/load-and-run/CMakeLists.txt

@@ -12,5 +12,10 @@ install (TARGETS load_and_run EXPORT ${MGE_EXPORT_TARGETS} RUNTIME DESTINATION $
 
 if(MGE_WITH_TEST)
     add_executable(json_loader_test test/json_loader_test.cpp src/json_loader.h src/json_loader.cpp)
+    # Windows does not support implicitly importing data members from DLL.
+    if (WIN32)
+        target_link_libraries (json_loader_test megbrain megdnn ${MGE_CUDA_LIBS})
+    else()
         target_link_libraries (json_loader_test megengine)
+    endif()
 endif()

src/CMakeLists.txt

@@ -197,7 +197,12 @@ endif()
 
 set (_VER_FILE ${PROJECT_SOURCE_DIR}/src/version.ld)
 
-if(MGE_BUILD_IMPERATIVE_RT
+# Windows does not support implicitly importing data members from DLL.
+# on Windows:
+# depends on megdnn/megbrain target, refs to sdk/load-and-run/CMakeLists.txt
+# depends on megengine lite_share or lite_static
+if(NOT WIN32)
+    if(MGE_BUILD_IMPERATIVE_RT
             )
         message(VERBOSE "create a export SHARED lib for python use")
         add_library(megengine_export SHARED)
@@ -207,23 +212,22 @@ if(MGE_BUILD_IMPERATIVE_RT
             message(VERBOSE "megengine_export configured to link megray")
             target_link_libraries(megengine_export PUBLIC megray)
         endif()
-endif()
-# Build as SHARED or STATIC depending on BUILD_SHARED_LIBS=ON/OFF
-add_library(megengine)
-target_link_libraries(megengine PRIVATE ${MGE_CUDA_LIBS})
-target_link_libraries(megengine PUBLIC megbrain megdnn)
-if (UNIX AND NOT APPLE)
+    endif()
+    # Build as SHARED or STATIC depending on BUILD_SHARED_LIBS=ON/OFF
+    add_library(megengine)
+    target_link_libraries(megengine PRIVATE ${MGE_CUDA_LIBS})
+    target_link_libraries(megengine PUBLIC megbrain megdnn)
+    if (UNIX AND NOT APPLE)
         target_link_options(megengine PRIVATE -Wl,--no-undefined -Wl,--version-script=${_VER_FILE})
         set_target_properties(megengine PROPERTIES LINK_DEPENDS ${_VER_FILE})
-endif()
-set_target_properties(megengine PROPERTIES CXX_VISIBILITY_PRESET default)
-set_target_properties(megengine PROPERTIES WINDOWS_EXPORT_ALL_SYMBOLS TRUE)
-# Do not export targets if MGE_WITH_DISTRIBUTED is on. MegRay is not ready
-# for this.
-install(TARGETS megengine
+    endif()
+    # Do not export targets if MGE_WITH_DISTRIBUTED is on. MegRay is not ready
+    # for this.
+    install(TARGETS megengine
         EXPORT ${MGE_EXPORT_TARGETS}
         LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
         ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR})
+endif()
 
 if (NOT MGE_WITH_DISTRIBUTED)
     install(TARGETS megbrain