c++ - 使用 OpenCv 2.4.10 + Cuda 7.5 + W10 时不支持 GPU

Question

我正在尝试使用以下方法启动并运行 GPU:

• OpenCv 2.4.10
• Visual Studio 2013
• Cuda 工具包 7.5(已测试并运行的 cuda 示例)
• Windows 10 x64
• 配备 NVIDIA GEFORCE GTX 860M 的联想 Y50

我使用带WITH_CUDA=ON的CMake编译了OpenCv，然后通过打开OpenCv.sln安装了库并构建了ALLBUILD和INSTALL项目。

将我的项目配置设置为:

• VC++ 目录

  • 包括目录:
    • C:\Program Files\NVIDIA GPU 计算工具包\CUDA\v7.5\include
    • C:\OpenCv\opencv\build\install\include(包含已编译的库)
    • C:\OpenCv\opencv\build\include

• 链接器-常规-附加库

  • C:\OpenCv\opencv\build\x64\vc12\lib
  • C:\OpenCv\opencv\build\install\x64\vc12\lib

• 链接器输入附加依赖项

  • opencv_calib3d2410d.lib
  • opencv_contrib2410d.lib
  • opencv_core2410d.lib
  • opencv_features2d2410d.lib
  • opencv_flann2410d.lib
  • opencv_gpu2410d.lib
  • opencv_highgui2410d.lib
  • opencv_imgproc2410d.lib
  • opencv_legacy2410d.lib
  • opencv_ml2410d.lib
  • opencv_nonfree2410d.lib
  • opencv_objdetect2410d.lib
  • opencv_ocl2410d.lib
  • opencv_photo2410d.lib
  • opencv_stitching2410d.lib
  • opencv_superres2410d.lib
  • opencv_ts2410d.lib
  • opencv_video2410d.lib
  • opencv_videostab2410d.lib

我的环境变量: 路径= C:\Program Files\NVIDIA GPU计算; Toolkit\CUDA\v7.5; C:\OpenCv\opencv\build\install\x64\vc12\bin; C:\OpenCv\opencv\build\x64\vc12\垃圾箱

测试代码:

   #include "opencv2\opencv.hpp"
   #include "opencv2\gpu\gpu.hpp"

    using namespace cv::gpu;
    using namespace std;

    int main() {
        {
        setDevice(0); 
        int gpu = getCudaEnabledDeviceCount();
        cout << gpu ;
        system("pause");
        return 0;
        }

}in

但是 setDevice 抛出

no GPU support (The library is compiled without CUDA support) in EmptyDeviceInfoFuncTable::setDevice, file c:\builds\2_4_PackSlave-win64-vc12-shared\opencv\dynamicuda\include\opencv2/dynamiccuda/dynamicuda.hpp. line 84

如果我注释 setDevice(0) 仍然未检测到 GPU，getCudaEnabledDeviceCount 返回 0。

Answer 1

问题是Opencv 2.4.10支持CUDA 4.2，而我的显卡仅兼容CUDA 6.5和7.5，我通过安装OpenCV 3.0.0解决了