[v1.8.0] Various CUDA 11.1 with BUILD_SPLIT_CUDA_FIXES (pytorch#52518)

Co-authored-by: Nikita Shulga <[email protected]>
Co-authored-by: peterjc123 <[email protected]>
Co-authored-by: Jane Xu <[email protected]>

Author: 4 people

caffe2/CMakeLists.txt

@@ -745,6 +745,10 @@ file(WRITE ${DUMMY_EMPTY_FILE} ${DUMMY_FILE_CONTENT})
 # Wrapper library for people who link against torch and expect both CPU and CUDA support
 # Contains "torch_cpu" and "torch_cuda"
 add_library(torch ${DUMMY_EMPTY_FILE})
+if(BUILD_SPLIT_CUDA)
+  # When we split torch_cuda, we want a dummy torch_cuda library that contains both parts
+  add_library(torch_cuda ${DUMMY_EMPTY_FILE})
+endif()
 if(HAVE_SOVERSION)
   set_target_properties(torch PROPERTIES
       VERSION ${TORCH_VERSION} SOVERSION ${TORCH_SOVERSION})
@@ -1237,34 +1241,39 @@ endif()
 
 caffe2_interface_library(torch_cpu torch_cpu_library)
 
-if(BUILD_SPLIT_CUDA)
-  caffe2_interface_library(torch_cuda_cu torch_cuda_cu_library)
-  caffe2_interface_library(torch_cuda_cpp torch_cuda_cpp_library)
-elseif(USE_CUDA)
+if(USE_CUDA)
   caffe2_interface_library(torch_cuda torch_cuda_library)
+  if(BUILD_SPLIT_CUDA)
+    caffe2_interface_library(torch_cuda_cu torch_cuda_cu_library)
+    caffe2_interface_library(torch_cuda_cpp torch_cuda_cpp_library)
+  endif()
 elseif(USE_ROCM)
   caffe2_interface_library(torch_hip torch_hip_library)
 endif()
 
 caffe2_interface_library(torch torch_library)
 
 install(TARGETS torch_cpu torch_cpu_library EXPORT Caffe2Targets DESTINATION "${TORCH_INSTALL_LIB_DIR}")
-if(BUILD_SPLIT_CUDA)
-  install(TARGETS torch_cuda_cu torch_cuda_cu_library EXPORT Caffe2Targets DESTINATION "${TORCH_INSTALL_LIB_DIR}")
-  install(TARGETS torch_cuda_cpp torch_cuda_cpp_library EXPORT Caffe2Targets DESTINATION "${TORCH_INSTALL_LIB_DIR}")
-elseif(USE_CUDA)
+
+if(USE_CUDA)
   install(TARGETS torch_cuda torch_cuda_library EXPORT Caffe2Targets DESTINATION "${TORCH_INSTALL_LIB_DIR}")
+  if(BUILD_SPLIT_CUDA)
+    install(TARGETS torch_cuda_cu torch_cuda_cu_library EXPORT Caffe2Targets DESTINATION "${TORCH_INSTALL_LIB_DIR}")
+    install(TARGETS torch_cuda_cpp torch_cuda_cpp_library EXPORT Caffe2Targets DESTINATION "${TORCH_INSTALL_LIB_DIR}")
+  endif()
 elseif(USE_ROCM)
   install(TARGETS torch_hip torch_hip_library EXPORT Caffe2Targets DESTINATION "${TORCH_INSTALL_LIB_DIR}")
 endif()
 install(TARGETS torch torch_library EXPORT Caffe2Targets DESTINATION "${TORCH_INSTALL_LIB_DIR}")
 
 target_link_libraries(torch PUBLIC torch_cpu_library)
-if(BUILD_SPLIT_CUDA)
-  target_link_libraries(torch PUBLIC torch_cuda_cu_library)
-  target_link_libraries(torch PUBLIC torch_cuda_cpp_library)
-elseif(USE_CUDA)
+
+if(USE_CUDA)
   target_link_libraries(torch PUBLIC torch_cuda_library)
+  if(BUILD_SPLIT_CUDA)
+    target_link_libraries(torch_cuda PUBLIC torch_cuda_cu_library)
+    target_link_libraries(torch_cuda PUBLIC torch_cuda_cpp_library)
+  endif()
 elseif(USE_ROCM)
   target_link_libraries(torch PUBLIC torch_hip_library)
 endif()

cmake/public/cuda.cmake

@@ -352,8 +352,13 @@ if(CAFFE2_STATIC_LINK_CUDA AND NOT WIN32)
       set_property(
         TARGET caffe2::cublas APPEND PROPERTY INTERFACE_LINK_LIBRARIES
         "${CUDA_TOOLKIT_ROOT_DIR}/lib64/libcublasLt_static.a")
+      # Add explicit dependency to cudart_static to fix
+      # libcublasLt_static.a.o): undefined reference to symbol 'cudaStreamWaitEvent'
+      # error adding symbols: DSO missing from command line
+      set_property(
+        TARGET caffe2::cublas APPEND PROPERTY INTERFACE_LINK_LIBRARIES
+        "${CUDA_cudart_static_LIBRARY}" rt dl)
     endif()
-    target_link_libraries(caffe2::cublas INTERFACE torch::cudart)
 else()
     set_property(
         TARGET caffe2::cublas PROPERTY INTERFACE_LINK_LIBRARIES

torch/utils/cpp_extension.py

@@ -22,16 +22,21 @@
 from setuptools.command.build_ext import build_ext
 from pkg_resources import packaging  # type: ignore
 
-BUILD_SPLIT_CUDA = os.getenv('BUILD_SPLIT_CUDA')
 IS_WINDOWS = sys.platform == 'win32'
 LIB_EXT = '.pyd' if IS_WINDOWS else '.so'
 EXEC_EXT = '.exe' if IS_WINDOWS else ''
+CLIB_PREFIX = '' if IS_WINDOWS else 'lib'
+CLIB_EXT = '.dll' if IS_WINDOWS else '.so'
 SHARED_FLAG = '/DLL' if IS_WINDOWS else '-shared'
 
 _HERE = os.path.abspath(__file__)
 _TORCH_PATH = os.path.dirname(os.path.dirname(_HERE))
 TORCH_LIB_PATH = os.path.join(_TORCH_PATH, 'lib')
 
+
+BUILD_SPLIT_CUDA = os.getenv('BUILD_SPLIT_CUDA') or (os.path.exists(os.path.join(
+    TORCH_LIB_PATH, f'{CLIB_PREFIX}torch_cuda_cu{CLIB_EXT}')) and os.path.exists(os.path.join(TORCH_LIB_PATH, f'{CLIB_PREFIX}torch_cuda_cpp{CLIB_EXT}')))
+
 # Taken directly from python stdlib < 3.9
 # See https://github.com/pytorch/pytorch/issues/48617
 def _nt_quote_args(args: Optional[List[str]]) -> List[str]: