在 QTCreator 中使用 MinGW 编译 c++ 和 cuda 代码

我正在尝试编译简单的cuda程序（我从在 Windows 上的 Qt Creator 中编译 Cuda 代码 https://stackoverflow.com/questions/12266264/compiling-cuda-code-in-qt-creator-on-windows） 这是我的 .pro 文件：

TARGET = Cuda

# Define output directories
DESTDIR = release
OBJECTS_DIR = release/obj
CUDA_OBJECTS_DIR = release/cuda

# Sourcefiles
SOURCES += main.cpp

# This makes the .cu files appear in your project
OTHER_FILES +=  vectorAddition.cu

# CUDA settings <-- may change depending on your system
CUDA_SOURCES += vectorAddition.cu

CUDA_SDK = "C:/Cuda/CudaSamples"   # Path to cuda SDK install
CUDA_DIR = "C:/Cuda/CudaToolkit"            # Path to cuda toolkit install
SYSTEM_NAME = Win32         # Depending on your system either 'Win32', 'x64', or 'Win64'
SYSTEM_TYPE = 32            # '32' or '64', depending on your system
CUDA_ARCH = sm_11           # Type of CUDA architecture, for example 'compute_10', 'compute_11', 'sm_10'
TEMP = 'C:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\bin' #tried to add VS compiler path, seems woring. When i changed it to higher version error appeared
VS = $$join(TEMP,'" -ccbin "','-ccbin "','"') #joing path to VS compiler with quotes
NVCC_OPTIONS = --use_fast_math $$VS

# include paths
INCLUDEPATH += $$CUDA_DIR/include \
               $$CUDA_SDK/common/inc/ \
               $$CUDA_SDK/../shared/inc/ \

# library directories
QMAKE_LIBDIR += $$CUDA_DIR/lib/$$SYSTEM_NAME \
                $$CUDA_SDK/common/lib/$$SYSTEM_NAME \
                $$CUDA_SDK/../shared/lib/$$SYSTEM_NAME \
                C:/CUDA/VS10/VC/lib \   #tried to add VS runtime libraries, after that some errors disappeared, and some new emerged so i guess it worked
QMAKE_LIBS += -lmsvcrt -llibcmt -llibcpmt #after adding VS runtime to QMAKE_LIBDIR this line seems to have no effect
# Add the necessary libraries
LIBS += -lcuda -lcudart
#LIBS += -lmsvcr100
QMAKE_CFLAGS_RELEASE += -v
MSVCRT_LINK_FLAG_DEBUG   = "/MTd"
MSVCRT_LINK_FLAG_RELEASE = "/MT"

# The following makes sure all path names (which often include spaces) are put between quotation marks
CUDA_INC = $$join(INCLUDEPATH,'" -I"','-I"','"')
#this was nested in CONFIG (see link above) but i dont think it worked
    cuda.input = CUDA_SOURCES
    cuda.output = $$CUDA_OBJECTS_DIR/${QMAKE_FILE_BASE}_cuda.o
    cuda.commands = $$CUDA_DIR/bin/nvcc.exe $$NVCC_OPTIONS $$CUDA_INC $$LIBS --machine $$SYSTEM_TYPE -arch=$$CUDA_ARCH \
                    -Xcompiler "/wd4819,/EHsc,/W3,/nologo,/O2,/Zi" \ 
                    -Xptxas -v \
                    -Xcompiler $$MSVCRT_LINK_FLAG_RELEASE \
     -c -o ${QMAKE_FILE_OUT} ${QMAKE_FILE_NAME} #Xptxas -v is command for some extra output during cuda compiling (nvvc). -Xcompiler, honesly i have no idea what "/wd4819,/EHsc,/W3,/nologo,/O2,/Zi" \ means,i copied those from some chinese webpage and again some errors went away
    cuda.dependency_type = TYPE_C
    QMAKE_EXTRA_COMPILERS += cuda

HEADERS += \
    vectorAddition.h #empty header, i was just trying something.

这是编译器输出：

13:09:17: Running steps for project Cuda...
13:09:17: Starting: "F:\Programs\Qt5.1.1\Tools\mingw48_32\bin\mingw32-make.exe" clean
F:/Programs/Qt5.1.1/Tools/mingw48_32/bin/mingw32-make -f Makefile.Release clean
mingw32-make[1]: Entering directory 'F:/workspaceSchool/build-Cuda-Desktop_Qt_5_1_1_MinGW_32bit-Release'
del release\cuda\vectorAddition_cuda.o
del release\cuda\vectorAddition_cuda.o release\obj\main.o
Could Not Find F:\workspaceSchool\build-Cuda-Desktop_Qt_5_1_1_MinGW_32bit-Release\release\cuda\vectorAddition_cuda.o
mingw32-make[1]: Leaving directory 'F:/workspaceSchool/build-Cuda-Desktop_Qt_5_1_1_MinGW_32bit-Release'
F:/Programs/Qt5.1.1/Tools/mingw48_32/bin/mingw32-make -f Makefile.Debug clean
mingw32-make[1]: Entering directory 'F:/workspaceSchool/build-Cuda-Desktop_Qt_5_1_1_MinGW_32bit-Release'
del release\cuda\vectorAddition_cuda.o
Could Not Find F:\workspaceSchool\build-Cuda-Desktop_Qt_5_1_1_MinGW_32bit-Release\release\cuda\vectorAddition_cuda.o
del release\cuda\vectorAddition_cuda.o debug\obj\main.o
Could Not Find F:\workspaceSchool\build-Cuda-Desktop_Qt_5_1_1_MinGW_32bit-Release\release\cuda\vectorAddition_cuda.o
Could Not Find F:\workspaceSchool\build-Cuda-Desktop_Qt_5_1_1_MinGW_32bit-Release\debug\obj\main.o
mingw32-make[1]: Leaving directory 'F:/workspaceSchool/build-Cuda-Desktop_Qt_5_1_1_MinGW_32bit-Release'
13:09:18: The process "F:\Programs\Qt5.1.1\Tools\mingw48_32\bin\mingw32-make.exe" exited normally.
13:09:18: Starting: "F:\Programs\Qt5.1.1\5.1.1\mingw48_32\bin\qmake.exe" F:\workspaceSchool\Cuda\Cuda.pro -r -spec win32-g++ -v
QMake version 3.0
Using Qt version 5.1.1 in F:\Programs\Qt5.1.1\5.1.1\mingw48_32\lib
13:09:18: The process "F:\Programs\Qt5.1.1\5.1.1\mingw48_32\bin\qmake.exe" exited normally.
13:09:18: Starting: "F:\Programs\Qt5.1.1\Tools\mingw48_32\bin\mingw32-make.exe" 
F:/Programs/Qt5.1.1/Tools/mingw48_32/bin/mingw32-make -f Makefile.Release
mingw32-make[1]: Entering directory 'F:/workspaceSchool/build-Cuda-Desktop_Qt_5_1_1_MinGW_32bit-Release'
C:\Cuda\CudaToolkit\bin\nvcc.exe --use_fast_math -ccbin "C:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\bin" -I"C:/Cuda/CudaToolkit/include" -I"C:/Cuda/CudaSamples/common/inc/" -I"C:/Cuda/CudaSamples/../shared/inc/" -lcuda -lcudart --machine 32 -arch=sm_11 -Xcompiler /wd4819,/EHsc,/W3,/nologo,/O2,/Zi -Xptxas -v -Xcompiler /MT -c -o release\cuda\vectorAddition_cuda.o ..\Cuda\vectorAddition.cu
ptxas : info : 0 bytes gmem
ptxas : info : Compiling entry function '__cuda_dummy_entry__' for 'sm_11'
ptxas : info : Used 0 registers
g++ -c -pipe -fno-keep-inline-dllexport -O2 -frtti -Wall -Wextra -fexceptions -mthreads -DUNICODE -DQT_NO_DEBUG -DQT_GUI_LIB -DQT_CORE_LIB -DQT_NEEDS_QMAIN -I..\Cuda -I"C:\Cuda\CudaToolkit\include" -I"C:\Cuda\CudaSamples\common\inc" -I"C:\Cuda\CudaSamples\..\shared\inc" -I"..\..\Programs\Qt5.1.1\5.1.1\mingw48_32\include" -I"..\..\Programs\Qt5.1.1\5.1.1\mingw48_32\include\QtGui" -I"..\..\Programs\Qt5.1.1\5.1.1\mingw48_32\include\QtCore" -I"release" -I"." -I"..\..\Programs\Qt5.1.1\5.1.1\mingw48_32\mkspecs\win32-g++" -o release\obj\main.o ..\Cuda\main.cpp
In file included from C:\Cuda\CudaToolkit\include/device_types.h:53:0,
                 from C:\Cuda\CudaToolkit\include/builtin_types.h:56,
                 from ..\Cuda\main.cpp:2:
C:\Cuda\CudaToolkit\include/host_defines.h:79:0: warning: "__cdecl" redefined [enabled by default]
 #define __cdecl
 ^
..\Cuda\main.cpp:1:0: note: this is the location of the previous definition
 #include <cuda.h>
 ^
In file included from C:\Cuda\CudaToolkit\include/builtin_types.h:60:0,
                 from ..\Cuda\main.cpp:2:
C:\Cuda\CudaToolkit\include/vector_types.h:73:0: warning: ignoring #pragma warning  [-Wunknown-pragmas]
 #pragma warning(push)
 ^
C:\Cuda\CudaToolkit\include/vector_types.h:74:0: warning: ignoring #pragma warning  [-Wunknown-pragmas]
 #pragma warning(disable: 4201 4408)
 ^
C:\Cuda\CudaToolkit\include/vector_types.h:350:0: warning: ignoring #pragma warning  [-Wunknown-pragmas]
 #pragma warning(pop)
 ^
g++ -Wl,-s -Wl,-subsystem,windows -mthreads -o release\Cuda.exe release/cuda/vectorAddition_cuda.o release/obj/main.o  -lglu32 -lopengl32 -lgdi32 -luser32 -lmingw32 -lqtmain -LC:\Cuda\CudaToolkit\lib\Win32 -LC:\Cuda\CudaSamples\common\lib\Win32 -LC:\Cuda\CudaSamples\..\shared\lib\Win32 -LC:\CUDA\VS10\VC\lib -LQMAKE_LIBS -L+= -L-lmsvcrt -L-llibcmt -L-llibcpmt -lcuda -lcudart -LF:\Programs\Qt5.1.1\5.1.1\mingw48_32\lib -lQt5Gui -lQt5Core 
Warning: .drectve `/FAILIFMISMATCH:"_MSC_VER=1600" /FAILIFMISMATCH:"_ITERATOR_DEBUG_LEVEL=0" /DEFAULTLIB:"libcpmt" /DEFAULTLIB:"LIBCMT" /DEFAULTLIB:"OLDNAMES" ' unrecognized
f:/programs/qt5.1.1/tools/mingw48_32/bin/../lib/gcc/i686-w64-mingw32/4.8.0/../../../../i686-w64-mingw32/lib/../lib/crt2.o:crtexe.c:(.text+0xcf): undefined reference to `_imp___fmode'
f:/programs/qt5.1.1/tools/mingw48_32/bin/../lib/gcc/i686-w64-mingw32/4.8.0/../../../../i686-w64-mingw32/lib/../lib/crt2.o:crtexe.c:(.text+0x2de): undefined reference to `_imp___acmdln'
f:/programs/qt5.1.1/tools/mingw48_32/bin/../lib/gcc/i686-w64-mingw32/4.8.0/../../../../i686-w64-mingw32/lib/../lib/crt2.o:crtexe.c:(.text+0x3cf): undefined reference to `_imp____initenv'
f:/programs/qt5.1.1/tools/mingw48_32/bin/../lib/gcc/i686-w64-mingw32/4.8.0/../../../../i686-w64-mingw32/bin/ld.exe: f:/programs/qt5.1.1/tools/mingw48_32/bin/../lib/gcc/i686-w64-mingw32/4.8.0/../../../../i686-w64-mingw32/lib/../lib/crt2.o: bad reloc address 0x0 in section `.CRT$XCAA'
collect2.exe: error: ld returned 1 exit status
Makefile.Release:78: recipe for target 'release\Cuda.exe' failed
mingw32-make[1]: Leaving directory 'F:/workspaceSchool/build-Cuda-Desktop_Qt_5_1_1_MinGW_32bit-Release'
makefile:34: recipe for target 'release' failed
mingw32-make[1]: *** [release\Cuda.exe] Error 1
mingw32-make: *** [release] Error 2
13:09:28: The process "F:\Programs\Qt5.1.1\Tools\mingw48_32\bin\mingw32-make.exe" exited with code 2.
Error while building/deploying project Cuda (kit: Desktop Qt 5.1.1 MinGW 32bit)
When executing step 'Make'
13:09:28: Elapsed time: 00:11.

以及所有问题（由于低代表而无法发布图像）：issues https://i.stack.imgur.com/2KRAw.png

如果我没有弄错的话，那么问题出在链接 cuda 目标文件（vectorAddition_cuda.o）上，该文件是由nvvc.exe with cl.exe以及 VS 库。然后MinGW(g++编译器）正在尝试以某种方式将 cuda 目标文件和 c++ 目标文件（main.o）与所有库链接到可执行文件中。然而，g++ 不知道 VS 运行时库并抛出这些错误。所以我想我需要以某种方式将 VS 库传递给 g++。 我怎样才能解决这个问题？

我有点困惑，你用的是MinGW还是Visual？标题似乎表明您正在使用 MinGW，但项目文件似乎混合使用了两者。你不能混合这两者。如果您使用 Visual Studio 2010 编译（或直接从 NVidia 下载二进制文件）CUDA，则必须使用 VS10 来编译您的项目，否则它将无法工作。

我自己从未使用过 CUDA，但似乎系统要求 http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-microsoft-windows/index.html仅提及 Visual Studio 2008、2010 和 2012。如果您想将其与 Qt 一起使用，这是可能的，您只需获取使用 VS 编译的 Qt（在下载页面 http://qt-project.org/downloads。您可以获得 Visual Studio Expressfor free http://www.visualstudio.com/en-us/downloads/download-visual-studio-vs#DownloadFamilies_2只要您不使用它创建任何商业应用程序。

要将 QtCreator 与 MSVC 后端编译器一起使用，请转到“工具”>“选项”>“构建和运行”>“套件”，然后添加一个带有 MSVC 编译器的新套件、cdb 作为调试器以及您刚刚下载的 Qt 版本（必须使用相同的 Visual Studio 进行编译） Studio 版本，否则将无法工作）。然后打开您的项目，转到“项目”选项卡（左侧）并选择您刚刚创建的套件。在一切顺利之前，您可能还应该清理您的 .pro 文件。

顺便说一句，您的链接器行中似乎有一些不合适的地方：

g++ -Wl,-s -Wl,-subsystem,windows -mthreads -o release\Cuda.exe release/cuda/vectorAddition_cuda.o release/obj/main.o  -lglu32 -lopengl32 -lgdi32 -luser32 -lmingw32 -lqtmain -LC:\Cuda\CudaToolkit\lib\Win32 -LC:\Cuda\CudaSamples\common\lib\Win32 -LC:\Cuda\CudaSamples\..\shared\lib\Win32 -LC:\CUDA\VS10\VC\lib -LQMAKE_LIBS -L+= -L-lmsvcrt -L-llibcmt -L-llibcpmt -lcuda -lcudart -LF:\Programs\Qt5.1.1\5.1.1\mingw48_32\lib -lQt5Gui -lQt5Core 

首先这个-L+=，这可能是由末尾的转义反斜杠引起的QMAKE_LIBDIR.

然后是语法-L-lmsvcrt似乎是错误的。可能是因为您正在使用QMAKE_LIBS，我个人从未使用过它，并且根据文档 http://qt-project.org/doc/qt-4.8/qmake-variable-reference.html#qmake-libs你也不应该这样做，因为它是一个内部变量。同样适用于QMAKE_LIBDIR http://qt-project.org/doc/qt-4.8/qmake-variable-reference.html#qmake-libdir顺便提一句。我只想用LIBS任何外部依赖项的变量。