I am using cmake (3.5.0) to compile a C++ probject on Ubuntu 14.04, which
depends on Nvidia Cudnn. It seems that the cmake does not provide the
correct paths of cudnn to the linker although I have configured the paths.
File FindCUDNN.cmake:
FIND_PATH(CUDNN_INCLUDE_DIR NAME "cudnn.h" PATHS "$ENV{CMAKE_INCLUDE_PATH}")
FIND_LIBRARY(CUDNN_LIBRARIES NAME "libcudnn.so" PATHS
"$ENV{CMAKE_LIBRARY_PATH}")
INCLUDE(FindPackageHandleStandardArgs)
find_package_handle_standard_args(CUDNN DEFAULT_MSG CUDNN_INCLUDE_DIR
CUDNN_LIBRARIES)
MARK_AS_ADVANCED(CUDNN_INCLUDE_DIR CUDNN_LIBRARIES)
CMakeLists.txt:
FIND_PACKAGE(CUDNN REQUIRED)
LIST(APPEND LINKER_LIBS ${CUDNN_LIBRARIES})
...
ADD_LIBRARY(myobjects OBJECT ${mysources})
ADD_LIBRARY(mylib SHARED $<TARGET_OBJECTS:myobjects> ${cuda_objs})
TARGET_LINK_LIBRARIES(mylib ${LINKER_LIBS})
CMAKE_INCLUDE_PATH, CMAKE_LIBRARY_PATH are exported to the paths of cudnn5
in my home folder. After cmake .., the paths in CmakeCache.txt are set to
cudnn5. However, the linker (and link.txt) uses -lcudnn without paths to
cudnn5. Consequently, it links the cudnn4 in the system folder, i.e.
/usr/local/cuda/lib64.
Note: If there is no cudnn4 in my system folder, everything is fine. The
link.txt file and the output of make VERBOSE=1 use the absolute paths of
cudnn 5 during linking.
Thanks.
Wei.
--
Powered by www.kitware.com
Please keep messages on-topic and check the CMake FAQ at:
http://www.cmake.org/Wiki/CMake_FAQ
Kitware offers various services to support the CMake community. For more
information on each offering, please visit:
CMake Support: http://cmake.org/cmake/help/support.html
CMake Consulting: http://cmake.org/cmake/help/consulting.html
CMake Training Courses: http://cmake.org/cmake/help/training.html
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
Follow this link to subscribe/unsubscribe:
http://public.kitware.com/mailman/listinfo/cmake
