Add option to build with integrated OpenCL (#3144)

* Add specialized OpenCL/Python package build path

* Refer to upstream OpenCL repository

* Reset build job count in setup.py

* TEMPORARY: refer to OpenCL fork to ensure Linux CI builds succeed

* Remove intermediate cmake target

* Restrict OpenCL headers to documented API version

* Use command line definition to activate integrated build

* Flag reference to unofficial repo with FIXME

* TEMPORARY: update private repo tag for dependency

* Remove integrated build for non-Win32 and related cleanup

* Remove commented code

* Rename integrated OpenCL build option and other cleanups

* Small cleanups

* Update CMakeIntegratedOpenCL.cmake
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* Update CMakeIntegratedOpenCL.cmake
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* Update CMakeIntegratedOpenCL.cmake
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* Update CMakeIntegratedOpenCL.cmake
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* Update CMakeLists.txt
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* Update CMakeLists.txt
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* Update CMakeLists.txt
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* Update CMakeIntegratedOpenCL.cmake

Targeted download of Boost submodules.
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* Update CMakeLists.txt
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

CMakeIntegratedOpenCL.cmake

+set(BUILD_SHARED_LIBS OFF CACHE BOOL "" FORCE)
+set(BOOST_VERSION_DOT "1.74")
+string(REPLACE "." "_" BOOST_VERSION_UNDERSCORE ${BOOST_VERSION_DOT})
+
+set(OPENCL_HEADER_REPOSITORY "https://github.com/KhronosGroup/OpenCL-Headers.git")
+set(OPENCL_HEADER_TAG "1b2a1850f410aaaaeaa56cead5a179b5aea4918e")
+
+set(OPENCL_LOADER_REPOSITORY "https://github.com/KhronosGroup/OpenCL-ICD-Loader.git")
+set(OPENCL_LOADER_TAG "98ca71fb9f8484f1cd1999f55224bf9e8d18693b")
+
+set(BOOST_REPOSITORY "https://github.com/boostorg/boost.git")
+set(BOOST_TAG "boost-${BOOST_VERSION_DOT}.0")
+
+# Build Independent OpenCL library
+include(FetchContent)
+FetchContent_Declare(OpenCL-Headers GIT_REPOSITORY ${OPENCL_HEADER_REPOSITORY} GIT_TAG ${OPENCL_HEADER_TAG})
+FetchContent_GetProperties(OpenCL-Headers)
+if(NOT OpenCL-Headers_POPULATED)
+  FetchContent_Populate(OpenCL-Headers)
+  message(STATUS "Populated OpenCL Headers")
+endif()
+set(OPENCL_ICD_LOADER_HEADERS_DIR ${opencl-headers_SOURCE_DIR} CACHE PATH "") # for OpenCL ICD Loader
+set(OpenCL_INCLUDE_DIR ${opencl-headers_SOURCE_DIR} CACHE PATH "") # for Boost::Compute
+
+FetchContent_Declare(OpenCL-ICD-Loader GIT_REPOSITORY ${OPENCL_LOADER_REPOSITORY} GIT_TAG ${OPENCL_LOADER_TAG})
+FetchContent_GetProperties(OpenCL-ICD-Loader)
+if(NOT OpenCL-ICD-Loader_POPULATED)
+  FetchContent_Populate(OpenCL-ICD-Loader)
+  set(USE_DYNAMIC_VCXX_RUNTIME ON)
+  add_subdirectory(${opencl-icd-loader_SOURCE_DIR} ${opencl-icd-loader_BINARY_DIR} EXCLUDE_FROM_ALL)
+  message(STATUS "Populated OpenCL ICD Loader")
+endif()
+list(APPEND INTEGRATED_OPENCL_INCLUDES ${OPENCL_ICD_LOADER_HEADERS_DIR})
+list(APPEND INTEGRATED_OPENCL_LIBRARIES ${opencl-icd-loader_BINARY_DIR}/Release/OpenCL.lib cfgmgr32.lib runtimeobject.lib)
+list(APPEND INTEGRATED_OPENCL_DEFINITIONS CL_TARGET_OPENCL_VERSION=120)
+
+# Build Independent Boost libraries
+include(ExternalProject)
+include(ProcessorCount)
+ProcessorCount(J)
+set(BOOST_BASE "${PROJECT_BINARY_DIR}/Boost")
+set(BOOST_BOOTSTRAP "${BOOST_BASE}/source/bootstrap.bat")
+set(BOOST_BUILD "${BOOST_BASE}/source/b2.exe")
+set(BOOST_FLAGS "")
+list(APPEND BOOST_SUBMODULES "libs/algorithm" "libs/align" "libs/any" "libs/array" "libs/assert" "libs/bind" "libs/chrono" "libs/compute" "libs/concept_check" "libs/config" "libs/container" "libs/container_hash" "libs/core" "libs/detail" "libs/filesystem" "libs/foreach" "libs/format" "libs/function" "libs/function_types" "libs/fusion" "libs/headers" "libs/integer" "libs/io" "libs/iterator" "libs/lexical_cast" "libs/math" "libs/move" "libs/mpl" "libs/multi_index" "libs/numeric/conversion" "libs/optional" "libs/predef" "libs/preprocessor" "libs/property_tree" "libs/range" "libs/ratio" "libs/serialization" "libs/smart_ptr" "libs/static_assert" "libs/system" "libs/throw_exception" "libs/tuple" "libs/typeof" "libs/type_index" "libs/type_traits" "libs/utility" "libs/uuid" "libs/winapi" "tools/boost_install" "tools/build")
+ExternalProject_Add(Boost
+  TMP_DIR "${BOOST_BASE}/tmp"
+  STAMP_DIR "${BOOST_BASE}/stamp"
+  DOWNLOAD_DIR "${BOOST_BASE}/download"
+  SOURCE_DIR "${BOOST_BASE}/source"
+  BINARY_DIR "${BOOST_BASE}/source"
+  INSTALL_DIR "${BOOST_BASE}/install"
+  GIT_REPOSITORY ${BOOST_REPOSITORY}
+  GIT_TAG ${BOOST_TAG}
+  GIT_SUBMODULES ${BOOST_SUBMODULES}
+  GIT_SHALLOW ON
+  UPDATE_COMMAND ""
+  PATCH_COMMAND ""
+  CONFIGURE_COMMAND ${BOOST_BOOTSTRAP}
+  BUILD_COMMAND ${BOOST_BUILD} -sBOOST_ROOT=${BOOST_BASE}/source -a -q -j ${J} --with-headers --with-chrono --with-filesystem --with-system link=static runtime-link=shared variant=release threading=multi cxxflags="${BOOST_FLAGS}"
+  INSTALL_COMMAND ""
+)
+set(BOOST_INCLUDE "${BOOST_BASE}/source" CACHE PATH "")
+set(BOOST_LIBRARY "${BOOST_BASE}/source/stage/lib" CACHE PATH "")
+list(APPEND INTEGRATED_OPENCL_INCLUDES ${BOOST_INCLUDE})
+if(MSVC)
+  if(${MSVC_VERSION} GREATER 1929)
+    message(FATAL_ERROR "Unrecognized MSVC version number: ${MSVC_VERSION}")
+  elseif(${MSVC_VERSION} GREATER 1919)
+    set(MSVC_TOOLCHAIN_ID "142")
+  elseif(${MSVC_VERSION} GREATER 1909)
+    set(MSVC_TOOLCHAIN_ID "141")
+  elseif(${MSVC_VERSION} GREATER 1899)
+    set(MSVC_TOOLCHAIN_ID "140")
+  else()
+    message(FATAL_ERROR "Unrecognized MSVC version number: ${MSVC_VERSION}")
+  endif()
+  list(APPEND INTEGRATED_OPENCL_LIBRARIES ${BOOST_LIBRARY}/libboost_filesystem-vc${MSVC_TOOLCHAIN_ID}-mt-x64-${BOOST_VERSION_UNDERSCORE}.lib)
+  list(APPEND INTEGRATED_OPENCL_LIBRARIES ${BOOST_LIBRARY}/libboost_system-vc${MSVC_TOOLCHAIN_ID}-mt-x64-${BOOST_VERSION_UNDERSCORE}.lib)
+  list(APPEND INTEGRATED_OPENCL_LIBRARIES ${BOOST_LIBRARY}/libboost_chrono-vc${MSVC_TOOLCHAIN_ID}-mt-x64-${BOOST_VERSION_UNDERSCORE}.lib)
+else()
+  message(FATAL_ERROR "Integrated OpenCL build is not yet available for MinGW")
+endif()
+
+set(BUILD_SHARED_LIBS ON CACHE BOOL "" FORCE)

CMakeLists.txt

-if(USE_GPU OR APPLE)
+if(__INTEGRATE_OPENCL)
+  cmake_minimum_required(VERSION 3.11)
+elseif(USE_GPU OR APPLE)
   cmake_minimum_required(VERSION 3.2)
 elseif(USE_CUDA)
   cmake_minimum_required(VERSION 3.16)
@@ -22,11 +24,18 @@ OPTION(USE_CUDA "Enable CUDA-accelerated training (EXPERIMENTAL)" OFF)
 OPTION(USE_DEBUG "Set to ON for Debug mode" OFF)
 OPTION(BUILD_STATIC_LIB "Build static library" OFF)
 OPTION(__BUILD_FOR_R "Set to ON if building lib_lightgbm for use with the R package" OFF)
+OPTION(__INTEGRATE_OPENCL "Set to ON if building LightGBM with the OpenCL ICD Loader and its dependencies included" OFF)
 
 if(APPLE)
     OPTION(APPLE_OUTPUT_DYLIB "Output dylib shared library" OFF)
 endif(APPLE)
 
+if(__INTEGRATE_OPENCL)
+  set(__INTEGRATE_OPENCL ON CACHE BOOL "" FORCE)
+  set(USE_GPU OFF CACHE BOOL "" FORCE)
+  message(STATUS "Building library with integrated OpenCL components")
+endif()
+
 if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
   if(CMAKE_CXX_COMPILER_VERSION VERSION_LESS "4.8.2")
     message(FATAL_ERROR "Insufficient gcc version")
@@ -134,6 +143,15 @@ if(USE_GPU)
     ADD_DEFINITIONS(-DUSE_GPU)
 endif(USE_GPU)
 
+if(__INTEGRATE_OPENCL)
+    if(WIN32)
+        include(CMakeIntegratedOpenCL.cmake)
+        ADD_DEFINITIONS(-DUSE_GPU)
+    else()
+        message(FATAL_ERROR "Integrated OpenCL build is available only for Windows")
+    endif(WIN32)
+endif(__INTEGRATE_OPENCL)
+
 if(USE_CUDA)
     find_package(CUDA REQUIRED)
     include_directories(${CUDA_INCLUDE_DIRS})
@@ -378,6 +396,19 @@ if(USE_GPU)
   TARGET_LINK_LIBRARIES(_lightgbm ${OpenCL_LIBRARY} ${Boost_LIBRARIES})
 endif(USE_GPU)
 
+if(__INTEGRATE_OPENCL)
+  # targets OpenCL and Boost are added in CMakeIntegratedOpenCL.cmake
+  add_dependencies(lightgbm OpenCL Boost)
+  add_dependencies(_lightgbm OpenCL Boost)
+  # variables INTEGRATED_OPENCL_* are set in CMakeIntegratedOpenCL.cmake
+  target_include_directories(lightgbm PRIVATE ${INTEGRATED_OPENCL_INCLUDES})
+  target_include_directories(_lightgbm PRIVATE ${INTEGRATED_OPENCL_INCLUDES})
+  target_compile_definitions(lightgbm PRIVATE ${INTEGRATED_OPENCL_DEFINITIONS})
+  target_compile_definitions(_lightgbm PRIVATE ${INTEGRATED_OPENCL_DEFINITIONS})
+  target_link_libraries(lightgbm PRIVATE ${INTEGRATED_OPENCL_LIBRARIES})
+  target_link_libraries(_lightgbm PRIVATE ${INTEGRATED_OPENCL_LIBRARIES})
+endif()
+
 if(USE_CUDA)
   set_target_properties(lightgbm PROPERTIES CUDA_RESOLVE_DEVICE_SYMBOLS ON)
   TARGET_LINK_LIBRARIES(
@@ -396,12 +427,14 @@ if(USE_HDFS)
   TARGET_LINK_LIBRARIES(_lightgbm ${HDFS_CXX_LIBRARIES})
 endif(USE_HDFS)
 
-if(WIN32 AND (MINGW OR CYGWIN))
+if(WIN32)
+    if(MINGW OR CYGWIN)
       TARGET_LINK_LIBRARIES(lightgbm Ws2_32)
       TARGET_LINK_LIBRARIES(_lightgbm Ws2_32)
       TARGET_LINK_LIBRARIES(lightgbm IPHLPAPI)
       TARGET_LINK_LIBRARIES(_lightgbm IPHLPAPI)
-endif()
+    endif(MINGW OR CYGWIN)
+endif(WIN32)
 
 if(__BUILD_FOR_R)
   if(MSVC)

python-package/setup.py

@@ -29,7 +29,7 @@ def find_lib():
     return LIB_PATH
 
 
-def copy_files(use_gpu=False):
+def copy_files(integrated_opencl=False, use_gpu=False):
 
     def copy_files_helper(folder_name):
         src = os.path.join(CURRENT_DIR, os.path.pardir, folder_name)
@@ -51,14 +51,18 @@ def copy_files(use_gpu=False):
         copy_file(os.path.join(CURRENT_DIR, os.path.pardir, "windows", "LightGBM.vcxproj"),
                   os.path.join(CURRENT_DIR, "compile", "windows", "LightGBM.vcxproj"),
                   verbose=0)
-        if use_gpu:
-            copy_files_helper('compute')
+        copy_file(os.path.join(CURRENT_DIR, os.path.pardir, "LICENSE"),
+                  os.path.join(CURRENT_DIR, "LICENSE"),
+                  verbose=0)
         copy_file(os.path.join(CURRENT_DIR, os.path.pardir, "CMakeLists.txt"),
                   os.path.join(CURRENT_DIR, "compile", "CMakeLists.txt"),
                   verbose=0)
-        copy_file(os.path.join(CURRENT_DIR, os.path.pardir, "LICENSE"),
-                  os.path.join(CURRENT_DIR, "LICENSE"),
+        if integrated_opencl:
+            copy_file(os.path.join(CURRENT_DIR, os.path.pardir, "CMakeIntegratedOpenCL.cmake"),
+                      os.path.join(CURRENT_DIR, "compile", "CMakeIntegratedOpenCL.cmake"),
                       verbose=0)
+        if use_gpu:
+            copy_files_helper('compute')
 
 
 def clear_path(path):
@@ -91,7 +95,7 @@ def compile_cpp(use_mingw=False, use_gpu=False, use_cuda=False, use_mpi=False,
                 use_hdfs=False, boost_root=None, boost_dir=None,
                 boost_include_dir=None, boost_librarydir=None,
                 opencl_include_dir=None, opencl_library=None,
-                nomp=False, bit32=False):
+                nomp=False, bit32=False, integrated_opencl=False):
 
     if os.path.exists(os.path.join(CURRENT_DIR, "build_cpp")):
         shutil.rmtree(os.path.join(CURRENT_DIR, "build_cpp"))
@@ -101,6 +105,9 @@ def compile_cpp(use_mingw=False, use_gpu=False, use_cuda=False, use_mpi=False,
     logger.info("Starting to compile the library.")
 
     cmake_cmd = ["cmake", "../compile/"]
+    if integrated_opencl:
+        use_gpu = False
+        cmake_cmd.append("-D__INTEGRATE_OPENCL=ON")
     if use_gpu:
         cmake_cmd.append("-DUSE_GPU=ON")
         if boost_root:
@@ -136,7 +143,7 @@ def compile_cpp(use_mingw=False, use_gpu=False, use_cuda=False, use_mpi=False,
         else:
             status = 1
             lib_path = os.path.join(CURRENT_DIR, "compile", "windows", "x64", "DLL", "lib_lightgbm.dll")
-            if not any((use_gpu, use_mpi, use_hdfs, nomp, bit32)):
+            if not any((use_gpu, use_mpi, use_hdfs, nomp, bit32, integrated_opencl)):
                 logger.info("Starting to compile with MSBuild from existing solution file.")
                 platform_toolsets = ("v142", "v141", "v140")
                 for pt in platform_toolsets:
@@ -189,6 +196,7 @@ class CustomInstall(install):
 
     user_options = install.user_options + [
         ('mingw', 'm', 'Compile with MinGW'),
+        ('integrated-opencl', None, 'Compile integrated OpenCL version'),
         ('gpu', 'g', 'Compile GPU version'),
         ('cuda', None, 'Compile CUDA version'),
         ('mpi', None, 'Compile MPI version'),
@@ -207,6 +215,7 @@ class CustomInstall(install):
     def initialize_options(self):
         install.initialize_options(self)
         self.mingw = 0
+        self.integrated_opencl = 0
         self.gpu = 0
         self.cuda = 0
         self.boost_root = None
@@ -231,12 +240,12 @@ class CustomInstall(install):
                                 "please use 64-bit Python instead.")
         open(LOG_PATH, 'wb').close()
         if not self.precompile:
-            copy_files(use_gpu=self.gpu)
+            copy_files(integrated_opencl=self.integrated_opencl, use_gpu=self.gpu)
             compile_cpp(use_mingw=self.mingw, use_gpu=self.gpu, use_cuda=self.cuda, use_mpi=self.mpi,
                         use_hdfs=self.hdfs, boost_root=self.boost_root, boost_dir=self.boost_dir,
                         boost_include_dir=self.boost_include_dir, boost_librarydir=self.boost_librarydir,
                         opencl_include_dir=self.opencl_include_dir, opencl_library=self.opencl_library,
-                        nomp=self.nomp, bit32=self.bit32)
+                        nomp=self.nomp, bit32=self.bit32, integrated_opencl=self.integrated_opencl)
         install.run(self)
         if os.path.isfile(LOG_PATH):
             os.remove(LOG_PATH)
@@ -245,7 +254,7 @@ class CustomInstall(install):
 class CustomSdist(sdist):
 
     def run(self):
-        copy_files(use_gpu=True)
+        copy_files(integrated_opencl=True, use_gpu=True)
         open(os.path.join(CURRENT_DIR, '_IS_SOURCE_PACKAGE.txt'), 'w').close()
         if os.path.exists(os.path.join(CURRENT_DIR, 'lightgbm', 'Release')):
             shutil.rmtree(os.path.join(CURRENT_DIR, 'lightgbm', 'Release'))