CMake-based build of CUDA app fails - no files passed to linker - c++

I'm trying to use CMake with a CUDA project of mine, but I'm having trouble getting it to build the executable when compiled on a system that has a CUDA-enabled device.
The CMakeLists.txt in question is below. It supports systems with and without CUDA-enabled devices, and builds just fine on my Macbook which doesn't have CUDA.
cmake_minimum_required (VERSION 2.8)
message(STATUS "CMake version: ${CMAKE_VERSION}")
project(stockModel)
# Grab the CUDA package
find_package(CUDA)
set(GPU_ACCELERATED ${CUDA_FOUND})
# Set directory and compilation flags for both g++ and nvcc
set(CMAKE_BINARY_DIR ${CMAKE_CURRENT_SOURCE_DIR})
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -pthread")
set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}
-gencode arch=compute_50,code=sm_50; -std=c++11; -lcurand;"
)
set(CUDA_PROPAGATE_HOST_FLAGS off)
# Add directories
include_directories(${CMAKE_CURRENT_SOURCE_DIR}/build/)
include_directories(${CMAKE_CURRENT_SOURCE_DIR}/core/)
if (${GPU_ACCELERATED})
include_directories(${CMAKE_CURRENT_SOURCE_DIR}/support/)
endif()
# Setup environments, depending on GPU accel. status
set(SRCS build/main.cpp core/callModels.cpp)
set(INCS core/callModels.h)
if (${GPU_ACCELERATED})
set(SRCS ${SRCS} support/prng.cu support/debugCFP.cu)
set(INCS ${INCS} support/prng.h support/debugCFP.h)
set_source_files_properties(${CMAKE_CURRENT_SOURCE_DIR}/core/callModels.cpp
PROPERTIES CUDA_SOURCE_PROPERTY_FORMAT OBJ
)
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}
-L/usr/local/cuda/lib64 -lcuda -lcudart"
)
endif()
# Create executable
message(STATUS "Sources: ${SRCS}")
message(STATUS "Includes: ${INCS}")
cuda_add_executable(stockModel ${SRCS} ${INCS})
The error I get when I attempt to build on my Jetson TX1 is as follows:
...
[ 80%] Building CXX object CMakeFiles/stockModel.dir/main.cpp.o
[100%] Linking CXX executable stockModel
c++: fatal error: no input files
compilation terminated.
...
Any ideas as to what is going wrong here? Obviously it has something to do with the CUDA 'extras', but I'm at a loss as to what is causing this.
Let me know if you need more details.
Here is the relevant part of the verbose output:
...
[100%] Linking CXX executable stockModel
/usr/local/bin/cmake -E cmake_link_script CMakeFiles/stockModel.dir/link.txt --verbose=1
/usr/bin/c++ -std=c++11 -pthread
c++: fatal error: no input files
compilation terminated.
I've uploaded the full make VERBOSE=1 output to this gist on GitHub.

CMake is sometimes finicky about spaces and list combinations. I know that doesn't sound like much of an explanation, but I'm not much of an expert.
What you need to do is replace this:
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}
-L/usr/local/cuda/lib64 -lcuda -lcudart"
with this:
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -L/usr/local/cuda/lib64 -lcuda -lcudart")
(single line). That should do it. At least - it does on my system (I created dummy source files with your files' names to try this out).

Related

C++: Problem with M1 architecture in cmake: how do I change the architecture?

I am trying to run cmake command in the new environment.
Upon running cmake . . I get clang: error: the clang compiler does not support '-march=core2' error. I tried to run it with cmake '-DCMAKE_CXX_FLAGS='-march=x86-64'' flag, and it changed the march for a while but only in CLion IDE and only when I was doing that continuously. I can't seem to make it work in the terminal while reading different makefile.
Full error when I run either cmake or ```cmake '-DCMAKE_CXX_FLAGS='-march=x86-64'':
Determining if the C compiler works failed with the following output:
Change Dir: /Users/marcin/CLionProjects/pythontest5/CMakeFiles/CMakeTmp
Run Build Command(s):/Library/Developer/CommandLineTools/usr/bin/make -f Makefile cmTC_925c1/fast && /Library/Developer/CommandLineTools/usr/bin/make -f CMakeFiles/cmTC_925c1.dir/build.make CMakeFiles/cmTC_925c1.dir/build
Building C object CMakeFiles/cmTC_925c1.dir/testCCompiler.c.o
/Library/Developer/CommandLineTools/usr/bin/gcc -march=core2 -mtune=haswell -mssse3 -ftree-vectorize -fPIC -fPIE -fstack-protector-strong -O2 -pipe -isystem /opt/homebrew/anaconda3/include -arch arm64 -isysroot /Library/Developer/CommandLineTools/SDKs/MacOSX.sdk -MD -MT CMakeFiles/cmTC_925c1.dir/testCCompiler.c.o -MF CMakeFiles/cmTC_925c1.dir/testCCompiler.c.o.d -o CMakeFiles/cmTC_925c1.dir/testCCompiler.c.o -c /Users/marcin/CLionProjects/pythontest5/CMakeFiles/CMakeTmp/testCCompiler.c
clang: error: the clang compiler does not support '-march=core2'
make[1]: *** [CMakeFiles/cmTC_925c1.dir/testCCompiler.c.o] Error 1
make: *** [cmTC_925c1/fast] Error 2
My CMakeLists.txt
cmake_minimum_required(VERSION 3.1)
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CXX_STANDARD_REQUIRED ON)
set(CMAKE_CXX_EXTENSIONS OFF)
project(automobile VERSION 0.1.0)
# Include dir
include_directories(/usr/local/include)
# Src
AUX_SOURCE_DIRECTORY(src SRC_FILES)
# Headers
set(PROJECT_SOURCE_DIR "src")
set(PROJECT_INCLUDE_DIR "include/automobile_bits")
# Source files
set(SOURCE_FILES
${PROJECT_INCLUDE_DIR}/motorcycle.hpp
${PROJECT_SOURCE_DIR}/motorcycle.cpp
)
# Set up such that XCode organizes the files correctly
source_group(TREE ${CMAKE_CURRENT_SOURCE_DIR} FILES ${SOURCE_FILES})
# Add library
add_library(automobile SHARED ${SOURCE_FILES})
# Include directories
target_include_directories(automobile PRIVATE include/automobile_bits)
# Install
install(TARGETS automobile DESTINATION lib)
# Install the headers
install(FILES include/automobile DESTINATION include)
# Create base directory
install(DIRECTORY include/automobile_bits DESTINATION include)
It's unclear where cmake gets -march=core2 from. Try to remove CMakeCache.txt then re-run cmake ..
If it does not help, try to remove CMakeCache.txt and run cmake . -DCMAKE_OSX_ARCHITECTURES=arm64.

ld cannot find -latomic or -lstdc++

When trying to compile my project using the toolchain provided by the manufacturer of the hardware I'm developing for, compilation succeeds, however linking fails with the following error messages:
/media/xxx/Data/toolchains/sysroots/x86_64-xxx-linux/usr/bin/arm-xxx-linux-gnueabi/../../libexec/arm-xxx-linux-gnueabi/gcc/arm-xxx-linux-gnueabi/4.8.4/ld: cannot find -latomic
/media/xxx/Data/toolchains/sysroots/x86_64-xxx-linux/usr/bin/arm-xxx-linux-gnueabi/../../libexec/arm-xxx-linux-gnueabi/gcc/arm-xxx-linux-gnueabi/4.8.4/ld: cannot find -lstdc++
However, when building a small test program using the same compiler and linker, everything works.
Below is an excerpt of the important parts of my CMake toolchain file:
set(CMAKE_SYSTEM_NAME Linux)
set(CMAKE_SYSTEM_PROCESSOR arm)
###
# Get $CROSS
###
if (NOT $ENV{CROSS_COMPILE} STREQUAL "")
set(cross $ENV{CROSS_COMPILE}) # WARNING: Ends with a - (minus)
else()
set(cross "arm-xxx-linux-gnueabi-")
endif()
###
# Set CMake sysroot
###
set(CMAKE_SYSROOT $ENV{SDKTARGETSYSROOT})
set(CMAKE_SYSROOT_LINK $ENV{SDKTARGETSYSROOT})
set(CMAKE_SYSROOT_COMPILE $ENV{SDKTARGETSYSROOT})
set(CMAKE_C_COMPILER $ENV{TOOLCHAIN_TOOLSDIR}/${cross}gcc)
set(CMAKE_CXX_COMPILER $ENV{TOOLCHAIN_TOOLSDIR}/${cross}g++)
set(CMAKE_LD $ENV{TOOLCHAIN_TOOLSDIR}/${cross}ld)
set(CMAKE_STRIP $ENV{TOOLCHAIN_TOOLSDIR}/${cross}strip)
set(CMAKE_AR $ENV{TOOLCHAIN_TOOLSDIR}/${cross}ar)
include_directories($ENV{SDKTARGETSYSROOT}/usr/include)
I'm not at liberty to disclose source code, but the source code isn't the issue here as it compiles correctly.
Here's another MWE of the CMakeLists.txt file:
cmake_minimum_required(VERSION 3.10)
project(xxx LANGUAGES CXX VERSION 2.0.0)
set(CMAKE_CXX_STANDARD 11)
set(CMAKE_CXX_STANDARD_REQUIRED True)
set(TARGET_NAME xxx.bin)
include_directories(
include/
${INCLUDE_DIRS}
)
file(GLOB_RECURSE FILES ${CMAKE_CURRENT_SOURCE_DIR} FOLLOW_SYMLINKS src/*.cpp)
if (xxx_DEBUG)
add_compile_options(
-Wpedantic
-DDDC_STUB
-fabi-version=6
-Wno-format-security
-Wno-reorder
-ggdb3
-O0
-fpermissive # Test
-fPIC
#-static-libstdc++ # with or without, doesn't change the behaviour
#-static-libgcc # with or without, doesn't change the behaviour
)
else()
add_compile_options(
-Wpedantic
-fabi-version=6
-Wno-format-security
-Wno-reorder
-O3
-fpermissive # Test
-fPIC
#-static-libstdc++ # with or without, doesn't change the behaviour
#-static-libgcc # with or without, doesn't change the behaviour
)
endif()
add_executable(${TARGET_NAME} ${FILES} ${MEMWATCH_FILES})
I've tried setting the sysroot-argument for ld, I've tried manually specifying the link (search) path -L$ENV{SDKTARGETSYSROOT}/usr/lib/.debug, and I've also tried hard-wiring the path to the library in question: $ENV{SDKTARGETSYSROOT}/usr/lib/.debug/libatomic.so.1.0.0
Is there another way I can tell ld what libraries to use?
This works perfectly with standard toolchains!
EDIT:
I've now also tried brute-forcing it by creating a symlink from the .debug/libatomic to usr/lib/libatomic. That also hasn't had the desired effect.

Is it possible to force CMake to reorder arguments when calling the linker? [duplicate]

This question already has an answer here:
Flag '-l' in CMAKE_CXX_FLAGS doesn't work
(1 answer)
Closed 4 years ago.
I have a strange behavior where CMake is not able to link any executable if dynamic linking is required. g++ is not able to find any of the symbols defined in the libraries.
I found out it might have to do with the order of the arguments CMake passes to g++ when linking.
Here is the verbose output of the build process (which fails linking):
[ 50%] Building CXX object CMakeFiles/chargen.dir/test.cpp.obj
g++ -std=c++11 -o CMakeFiles/test.dir/test.cpp.obj -c /f/test/test.cpp
[100%] Linking CXX executable mytest
/usr/bin/cmake -E cmake_link_script CMakeFiles/test.dir/link.txt --verbose=1
g++ -std=c++11 -lSDL2 CMakeFiles/test.dir/test.cpp.obj -o mytest
Indeed if I try to link using that command I get undefined references, however if I compile with the library flags at the end:
g++ -std=c++11 CMakeFiles/test.dir/test.cpp.obj -o mytest -lSDL2
it runs just fine. How can I force CMake to use that order of arguments instead?
The contents of the CMakeLists.txt:
cmake_minimum_required(VERSION 3.6)
project(mytest)
set(COMPILER_PATH "/usr/bin/")
set(CMAKE_MAKE_PROGRAM "make")
set(CMAKE_C_COMPILER "gcc")
set(CMAKE_CXX_COMPILER "g++")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -lSDL2")
set(SOURCE_FILES test.cpp)
add_executable(mytest ${SOURCE_FILES})
You have probably added -lSDL2 into CMAKE_CXX_FLAGS. In that case it will appear withn compile/link flags, in front of source file name in actual link command, as you showed. Don't do that.
You should use _target_link_libraries_ cmake command in order to define libraries to link against. In brief, the skeleton of your CMakeLists.txt should look like this:
project (project_name)
add_executable (mytest test.cpp)
target_link_libraries( mytest SDL2)
In this case, cmake will put libraries on correct place. Note that you may use target_link_libraries after adding target mytest
[EDIT]
After seeing your CMakeLists it is obvious that your problem is in wrong CMAKE_EXE_LINKER_FLAGS. Simply delete the line where you are setting it, and add target_link_libraries after add_executable.
Regarding your question about handling circular dependencies among static libs, you handle it the sam way as you would if you didn't use cmake: mention it twice:
target_link_libraries(
mytest
circular_dep1
circular_dep2
circular_dep1
circular_dep2
)
regarding your question about specific linker flags, just place them into target_link_libraries command, it accepts linker flags as well. You can find documentation on following link:
https://cmake.org/cmake/help/latest/command/target_link_libraries.html

cross compile dll on mac OSX with cmake and mingw

I installed the bwapi via wine on Mac OSX. And I try to compile the ExampleAIModule in it with CMake and MinGW on Mac OSX. The ExampleAIModule is actually a VS2017 project, I write a CMakeLists.txt, ran cmake CMakeLists.txt && make VERBOSE=1 got some errors. I want to know how to fix my CMakeLists.txt?
The folder structure:
BWAPI/ExampleAIModule/CMakeLists.txt
BWAPI/ExampleAIModule/Source/*.cpp *.h
BWAPI/include/*.h
BWAPI/lib/BWAPI.lib
BWAPI/lib/BWAPIClient.lib
make got errors:
Linking CXX shared library libExampleAIModule.dylib
/usr/local/Cellar/cmake/3.2.2/bin/cmake -E cmake_link_script CMakeFiles/ExampleAIModule.dir/link.txt --verbose=1
i686-w64-mingw32-g++ -std=c++11 -O3 -DNDEBUG -dynamiclib -Wl,-headerpad_max_install_names -o libExampleAIModule.dylib -install_name #rpath/libExampleAIModule.dylib CMakeFiles/ExampleAIModule.dir/Source/Dll.cpp.o CMakeFiles/ExampleAIModule.dir/Source/ExampleAIModule.cpp.o -L/.wine/drive_c/Starcraft/BWAPI/ExampleAIModule/../lib
i686-w64-mingw32-g++: error: rpath/libExampleAIModule.dylib: No such file or directory
i686-w64-mingw32-g++: error: unrecognized command line option ‘-install_name’
make[2]: *** [libExampleAIModule.dylib] Error 1
make[1]: *** [CMakeFiles/ExampleAIModule.dir/all] Error 2
CMakeLists.txt
cmake_minimum_required(VERSION 3.0.0 FATAL_ERROR)
################### Variables. ####################
# Change if you want modify path or other values. #
###################################################
set(PROJECT_NAME ExampleAIModule)
# Output Variables
# Folders files
set(CPP_DIR_1 Source)
set(HEADER_DIR_1 Source)
############## Define Project. ###############
# ---- This the main options of project ---- #
##############################################
project(${PROJECT_NAME} CXX)
# Define Release by default.
if(NOT CMAKE_BUILD_TYPE)
set(CMAKE_BUILD_TYPE "Release")
message(STATUS "Build type not specified: defaulting to release.")
endif(NOT CMAKE_BUILD_TYPE)
# Definition of Macros
add_definitions(
-DNOMINMAX
-D_DEBUG
-D_WINDOWS
-D_USRDLL
-DEXAMPLEAIMODULE_EXPORTS
-DUNICODE
-D_UNICODE
)
################# Flags ################
# Defines Flags for Windows and Linux. #
########################################
if(MSVC)
set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} /W3 /EHsc")
set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /W3 /EHsc")
endif(MSVC)
if(NOT MSVC)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
if ("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang")
set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -stdlib=libc++")
endif()
endif(NOT MSVC)
########### Add by me #############
SET(CMAKE_C_COMPILER i686-w64-mingw32-gcc)
SET(CMAKE_CXX_COMPILER i686-w64-mingw32-g++)
include_directories(../include)
link_directories(../lib)
################ Files ################
# -- Add files to project. -- #
#######################################
file(GLOB SRC_FILES
${CPP_DIR_1}/*.cpp
${HEADER_DIR_1}/*.h
)
# Add library to build.
add_library(${PROJECT_NAME} SHARED
${SRC_FILES}
)
CMake always use toolchains to compile for a specific platform/target, as mentioned here. However, for cross-compilation, you need to specify the toolchain file it needs to use. This is done using the CMAKE_TOOLCHAIN_FILE variable, as explained here.
The parameters related to the compilation (such as the system name, processor, compiler, etc.) on the target platform shouldn't be in the CMakeLists.txt file where you defined your project (using the project function).
Instead, you're supposed to set the compilers and other parameters related to cross-compilation in a toolchain file. And you may have a toolchain file for each platform you like to cross-compile for.
This is how your toolchain file may look like, assuming that the compilers are in the path:
set(CMAKE_SYSTEM_NAME Windows)
set(CMAKE_SYSTEM_PROCESSOR x86_64)
SET(CMAKE_C_COMPILER i686-w64-mingw32-gcc)
SET(CMAKE_CXX_COMPILER i686-w64-mingw32-g++)
If your toolchain file has the path /toolchains/windows.cmake, you need to pass -DCMAKE_TOOLCHAIN_FILE=/toolchains/windows.cmake to cmake.
This design keeps your project's CMakeLists.txt file toolchain-agnostic.

nvlink error when linking CUDA code against CUDA static library - CMake [duplicate]

The buildtime of my cuda library is increasing and so I thought that separate compilation introduced in CUDA 5.0 might help me. I couldn't figure out how to achieve separate compilation with cmake. I looked into the NVCC documentation and found how to compile device object (using the -dc option) and how to link them (using the -dlink). My attempts to get it running using cmake failed. I'm using cmake 2.8.10.2 and the head of the trunk of the FindCUDA.cmake. I couldn't however find out how to specify which files should be compiled and how to link them into a library.
Especially the syntax of the function(CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS output_file_var cuda_target options object_files source_files)
is unclear to me because I don't know what the output_file_var and the cuda_target are.
Here the not working results of my attemps:
cuda_compile(DEVICEMANAGER_O devicemanager.cu OPTIONS -dc)
cuda_compile(BLUB_O blub.cu OPTIONS -dc)
CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS(TEST_O gpuacceleration
"" DEVICEMANGER_O BLUB_O)
set(LIB_TYPE SHARED)
#cuda_add_library(gpuacceleration ${LIB_TYPE}
#${gpuacc_SRCS}
#devicemanager.cu
# blub.cu
#DEVICEMANAGER_O
# TEST_O
#)
Does anyone know how to compile and link a cuda library using cmake?
Thanks in advance.
EDIT:
After a friend consulted the developer of the FindCUDA.cmake, a bug got fixed in the example provided with FindCUDA.cmake (https://gforge.sci.utah.edu/gf/project/findcuda/scmsvn/?action=browse&path=%2Fcheckout%2Ftrunk%2FFindCuda.html).
I'm now able to build the example.
In my project I can build the library as needed using the following (cmake 2.8.10 required):
set(LIB_TYPE SHARED)
set(CUDA_SEPARABLE_COMPILATION ON)
cuda_add_library(gpuacceleration ${LIB_TYPE}
blub.cu
blab.cu
)
BUT:
I cannot link against this library. When I builded the lib without separate compilation i was able to link against it.
Now getting the following error:
undefined reference to `__cudaRegisterLinkedBinary_53_tmpxft_00005ab4_00000000_6_blub_cpp1_ii_d07d5695'
for every file with a function used in the interface. Seems strange since it builds without any warning etc.
Any ideas how to get this working?
EDIT:
I finally figured out how to do this. See #PHD's and my answer for details.
I finally got it running ;)
In Addition to the answer of #PHD and my comment on it I modified: set(BUILD_SHARED_LIBS OFF) in my CMakeLists.txt since shared libs are not supported for separate compilation according to the nvcc manually v5.0 page 40.
In addition to that use the latest rev (1223) from the repository instead of rev 1221. I contacted the developer and he fixed some issue blocking this.
This revision doesn't set the nvcc -arch=sm_xx flag correctly, so I added this manually for my project and informed the developer of FindCUDA.cmake. So this might get fixed in the future.
Don't forget to get cmake > 2.8.10 for this to work.
Hope this helps anyone but me ;)
Here is my CMakeLists.txt:
#Required for CUDA-Check
cmake_minimum_required(VERSION 2.8.10)
project(gpulib)
set(CMAKE_MODULE_PATH "${CMAKE_SOURCE_DIR}/CMake/cuda" ${CMAKE_MODULE_PATH})
# ============================================
# === Target
# ============================================
file(GLOB_RECURSE gpuacc_SRCS "*.cu")
include_directories(.)
# ---------------------------------------
# Find Cuda
find_package(CUDA REQUIRED)
set(CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE ON)
set(BUILD_SHARED_LIBS OFF)
set(CUDA_SEPARABLE_COMPILATION ON)
#list(APPEND CUDA_NVCC_FLAGS -arch=sm_20)
set(LIB_NAME "gpuacceleration")
cuda_add_library(${LIB_NAME}
${gpuacc_SRCS}
OPTIONS -DSTUFF="blah blah"
RELEASE -DNDEBUG
DEBUG -g -DDEBUG
)
set(PUBLIC_HEADERS "myheader1.h;myheader2.h")
INSTALL(FILES ${PUBLIC_HEADERS} DESTINATION include)
INSTALL(FILES "${CMAKE_BINARY_DIR}/src/lib${LIB_NAME}.a" DESTINATION lib)
EDIT: this is not working!
The problem is that there are undefined references to all cuda functions (eg. cudaMalloc) when linking the generated library when building a executable in the main project.
Still working on it
EDIT (2016-03-15): Yes, it is confirmed as a bug in FindCUDA: https://cmake.org/Bug/view.php?id=15157
TL;DR: This seems to be a bug in FindCUDA, which makes objects lose info on external definitions before the final linking.
The problem is that, even if separable compilation is enabled, a linking step is still performed for all the targets individually before the final linking.
For instance, I have module.cu with:
#include "module.h"
#include <cstdio>
double arr[10] = {1,2,3,4,5,6,7,8,9,10};
__constant__ double carr[10];
void init_carr() {
cudaMemcpyToSymbol(carr,arr,10*sizeof(double));
}
__global__ void pkernel() {
printf("(pkernel) carr[%d]=%g\n",threadIdx.x,carr[threadIdx.x]);
}
void print_carr() {
printf("in print_carr\n");
pkernel<<<1,10>>>();
}
and module.h with:
extern __constant__ double carr[10];
extern double arr[10];
void print_carr();
void init_carr();
and finally main.cu with:
#include "module.h"
#include <cstdio>
__global__ void kernel() {
printf("(kernel) carr[%d]=%g\n",threadIdx.x,carr[threadIdx.x]);
}
int main(int argc, char *argv[]) {
printf("arr: %g %g %g ..\n",arr[0],arr[1],arr[2]);
kernel<<<1,10>>>();
cudaDeviceSynchronize();
print_carr();
cudaDeviceSynchronize();
init_carr();
cudaDeviceSynchronize();
kernel<<<1,10>>>();
cudaDeviceSynchronize();
print_carr();
cudaDeviceSynchronize();
return 0;
}
This then works fine with the following Makefile:
NVCC=nvcc
NVCCFLAGS=-arch=sm_20
LIB=libmodule.a
OBJS=module.o main.o
PROG=extern
$(PROG): main.o libmodule.a
$(NVCC) $(NVCCFLAGS) -o $# $^
%.o: %.cu
$(NVCC) $(NVCCFLAGS) -dc -c -o $# $^
$(LIB): module.o
ar cr $# $^
clean:
$(RM) $(PROG) $(OBJS) $(LIB)
But then I try to use the following CMakeLists.txt:
CMAKE_MINIMUM_REQUIRED(VERSION 2.8.8)
PROJECT(extern)
FIND_PACKAGE(CUDA REQUIRED)
SET(CUDA_SEPARABLE_COMPILATION ON)
SITE_NAME(HOSTNAME)
SET(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -arch=sm_20)
cuda_add_library(module module.cu)
CUDA_ADD_EXECUTABLE(extern main.cu)
TARGET_LINK_LIBRARIES(extern module)
When then compiling, what then happens is that the following:
$ cmake ..
-- The C compiler identification is GNU 4.9.2
...
$ make VERBOSE=1
...
[ 25%] Building NVCC (Device) object CMakeFiles/module.dir//./module_generated_module.cu.o
...
-- Generating <...>/build/CMakeFiles/module.dir//./module_generated_module.cu.o
/usr/local/cuda/bin/nvcc <...>/module.cu -dc -o <...>/build/CMakeFiles/module.dir//./module_generated_module.cu.o -ccbin /usr/bin/cc -m64 -Xcompiler ,\"-g\" -arch=sm_20 -DNVCC -I/usr/local/cuda/include
[ 50%] Building NVCC intermediate link file CMakeFiles/module.dir/./module_intermediate_link.o
/usr/local/cuda/bin/nvcc -arch=sm_20 -m64 -ccbin "/usr/bin/cc" -dlink <...>/build/CMakeFiles/module.dir//./module_generated_module.cu.o -o <...>/build/CMakeFiles/module.dir/./module_intermediate_link.o
...
/usr/bin/ar cr libmodule.a CMakeFiles/module.dir/./module_generated_module.cu.o CMakeFiles/module.dir/./module_intermediate_link.o
/usr/bin/ranlib libmodule.a
...
[ 50%] Built target module
[ 75%] Building NVCC (Device) object CMakeFiles/extern.dir//./extern_generated_main.cu.o
...
-- Generating <...>/build/CMakeFiles/extern.dir//./extern_generated_main.cu.o
/usr/local/cuda/bin/nvcc <...>/main.cu -dc -o <...>/build/CMakeFiles/extern.dir//./extern_generated_main.cu.o -ccbin /usr/bin/cc -m64 -Xcompiler ,\"-g\" -arch=sm_20 -DNVCC -I/usr/local/cuda/include -I/usr/local/cuda/include
...
[100%] Building NVCC intermediate link file CMakeFiles/extern.dir/./extern_intermediate_link.o
/usr/local/cuda/bin/nvcc -arch=sm_20 -m64 -ccbin "/usr/bin/cc" -dlink <...>/build/CMakeFiles/extern.dir//./extern_generated_main.cu.o -o <...>/build/CMakeFiles/extern.dir/./extern_intermediate_link.o
nvlink error : Undefined reference to 'carr' in '<...>/build/CMakeFiles/extern.dir//./extern_generated_main.cu.o'
Clearly, the problem are the nvcc -dlink obj.o -o obj_intermediate_link.o lines. Then, I guess, the info on external definitions are lost. So, the question is, it is possible to make CMake/FindCUDA not do this extra linking step?
Otherwise, I would argue that this is a bug. Do you agree? I can file a bug report with CMake.
Tested it with nvcc version:
nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2012 NVIDIA
Corporation Built on Fri_Sep_21_17:28:58_PDT_2012 Cuda compilation
tools, release 5.0, V0.2.1221
and svn revision:
URL: https://gforge.sci.utah.edu/svn/findcuda/trunk
Repository Root: https://gforge.sci.utah.edu/svn/findcuda
Repository UUID: 81322f20-870f-0410-845c-a4cd4664c908
Revision: 1221
Node Kind: directory
Schedule: normal
Last Changed Rev: 1221
Last Changed Date: 2013-01-28 22:31:07 +0100 (Mon, 28 Jan 2013)
In this example includes following classes:
lib.h / lib.cu
kernel.h / kernel.cu
kernel.cu contains a simple CUDA kernel and a class with a public method to call the CUDA kernel. The class lib contains an instance of the class kernel and a method calling the public method of class kernel.
Following CMakeLists.txt works with this configuration:
cmake_minimum_required(VERSION 2.6.2)
project(Cuda-project)
set(CMAKE_MODULE_PATH "${CMAKE_SOURCE_DIR}/CMake/cuda" ${CMAKE_MODULE_PATH})
find_package(CUDA QUIET REQUIRED)
set(CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE OFF)
set(BUILD_SHARED_LIBS ON)
list(APPEND CUDA_NVCC_FLAGS -DBLAH="he he" -DTEST1="this is a test")
CUDA_ADD_LIBRARY(test_lib
kernel.cu
lib.cu
# SHARED
# STATIC
OPTIONS -DSTUFF="blah blah"
RELEASE --use_fast_math -DNDEBUG
DEBUG -g -DDEBUG
)
INSTALL(FILES lib.h kernel.h
DESTINATION include)
INSTALL(FILES "${CMAKE_BINARY_DIR}/libtest_lib.so"
DESTINATION lib)
I couldn't make it works using CUDA_ADD_EXECUTABLE so I created a function that makes a custom target to do so.
function(add_cuda_exe_lib name files libraries is_lib)
set (obj_list)
foreach(file ${files})
add_custom_command(
OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/${file}.o
DEPENDS ${file}
COMMAND ${CMAKE_COMMAND} -E cmake_echo_color --blue "Building NVCC Device object ${CMAKE_CURRENT_SOURCE_DIR}/${file}"
COMMAND ${CUDA_NVCC_EXECUTABLE} ${CUDA_NVCC_FLAGS} -dc "${CMAKE_CURRENT_SOURCE_DIR}/${file}" -o "${CMAKE_CURRENT_BINARY_DIR}/${file}.o"
COMMENT "Building ${CMAKE_CURRENT_SOURCE_DIR}/${file}"
VERBATIM
)
LIST(APPEND obj_list ${CMAKE_CURRENT_BINARY_DIR}/${file}.o)
endforeach()
set (lib_list)
LIST(APPEND lib_list "-lcudadevrt")
foreach(library_name ${libraries})
LIST(APPEND lib_list "-l${library_name}")
endforeach()
set (flags ${CUDA_NVCC_FLAGS})
if (is_lib)
LIST(APPEND flags "-dlink")
set (obj_name "${CMAKE_CURRENT_BINARY_DIR}/${name}.so")
else()
set (obj_name "${CMAKE_CURRENT_BINARY_DIR}/${name}")
endif()
add_custom_target(${name} ALL
COMMAND ${CMAKE_COMMAND} -E cmake_echo_color --red "Linking CXX executable ${name}"
COMMAND ${CUDA_NVCC_EXECUTABLE} ${flags} ${obj_list} ${lib_list} -o ${obj_name}
DEPENDS ${obj_list}
COMMENT "Linking ${name}"
)
endfunction()
function(add_cuda_exe name files libraries)
add_cuda_exe_lib(${name} "${files}" "${libraries}" OFF)
endfunction()
function(add_cuda_lib name files libraries)
add_cuda_exe_lib(${name} "${files}" "${libraries}" ON)
endfunction()
Now, to generate a lib, just use:
add_cuda_lib(testar "devicemanager.cu;blub.cu" "")
Or this to generate an executable:
add_cuda_exe(testar "devicemanager.cu;blub.cu" "")
The last param is a list of libs to be attached.
I hope it helps.