库已链接但引用未定义

Question

我正在尝试使用以前工作过一次的 NVIDIA 卡在 Ubuntu 上编译一个 openCL 程序，

#include <CL/cl.h>
#include <iostream>
#include <vector>

using namespace std;

int main() {
  cl_platform_id platform;
  cl_device_id device;
  cl_context context;
  cl_command_queue command_queue;
  cl_int error;

  if(clGetPlatformIDs(1, &platform, NULL) != CL_SUCCESS) {
    cout << "platform error" << endl;
  }

  if(clGetDeviceIDs(platform, CL_DEVICE_TYPE_GPU, 1, &device, NULL) != CL_SUCCESS) {
    cout << "device error" << endl;
  }

  context = clCreateContext(NULL, 1, &device, NULL, NULL, &error);
  if(error != CL_SUCCESS) {
    cout << "context error" << endl;
  }

  command_queue = clCreateCommandQueue(context, device, 0, &error);
  if(error != CL_SUCCESS) {
    cout << "command queue error" << endl;
  }

  return 0;
}

我是这样编译的，

g++ -I/usr/local/cuda/include -L/usr/lib/nvidia-current -lOpenCL opencl.cpp

我得到了这个结果

/tmp/ccAdS9ig.o: In function `main':
opencl.cpp:(.text+0x1a): undefined reference to `clGetPlatformIDs'
opencl.cpp:(.text+0x3d): undefined reference to `clGetDeviceIDs'
opencl.cpp:(.text+0x65): undefined reference to `clCreateContext'
opencl.cpp:(.text+0x85): undefined reference to `clCreateCommandQueue'
collect2: ld returned 1 exit status

但是

nm -D /usr/lib/nvidia-current/libOpenCL.so

告诉我libOpenCL.so至少包含clGetPlatformIDs

0000000000002400 T clGetKernelWorkGroupInfo
0000000000002140 T clGetMemObjectInfo
0000000000002e80 T clGetPlatformIDs
0000000000002de0 T clGetPlatformInfo
0000000000002310 T clGetProgramBuildInfo
00000000000022f0 T clGetProgramInfo
00000000000021f0 T clGetSamplerInfo

我错过了什么吗？

Answer 1

链接时，库和源文件的顺序会有所不同。例如你的情况，

g++ -I/usr/local/cuda/include -L/usr/lib/nvidia-current -lOpenCL opencl.cpp

OpenCL 库中定义的函数可能不会加载，因为在它们要求查找之前没有任何内容。但是如果你使用，

g++ opencl.cpp -I/usr/local/cuda/include -L/usr/lib/nvidia-current -lOpenCL  

然后任何函数请求都将在 OpenCL 库中找到并加载它们。

Answer 2

来自

gcc

手册页：

   -llibrary
   -l library
       Search the library named library when linking.  (The second alternative with the library as a separate argument is only for POSIX compliance and is not recommended.)

       It makes a difference where in the command you write this option; the linker searches and processes libraries and object files in the order they are specified.  Thus, foo.o
       -lz bar.o searches library z after file foo.o but before bar.o.  If bar.o refers to functions in z, those functions may not be loaded.

       The linker searches a standard list of directories for the library, which is actually a file named liblibrary.a.  The linker then uses this file as if it had been specified
       precisely by name.

因此，请尝试在编译命令中的文件参数后指定

-lOpenCL

。

您还可以在共享库文件 libOpenCL.so 中搜索符号。使用您的命令，您可以将程序再次链接到静态库，格式为

libOpenCL.a

.

Answer 3

或者，您可以将标头和库路径添加到全局变量中。

export CPLUS_INCLUDE_PATH=$CPLUS_INCLUDE_PATH:/usr/local/cuda/include
export LIBRARY_PATH=$LIBRARY_PATH:/usr/lib/nvidia-current

你也可以尝试设置

export PATH=$PATH:/usr/local/cuda/bin

现在应该可以跑了

g++ opencl.cpp

Answer 4

就我而言，我编译了一个 C++ 应用程序，但链接了一个 C 库。该 lib 包含的头文件未将函数原型指定为 extern "C"，因此链接器正在搜索经过修饰的函数名称而不是纯 C 名称。 在 C 标头周围指定 extern "C" 包括为我解决了这个问题。