opencl - 如何使用 opencl 在内核中显示图像？

Question

我是opencl的新手。任务是：

加载预先存在的图像
使用opencl编写主机代码将图像ptr发送到内核
计算内核中加载图像的hsl阈值
显示阈值或二值图像

我已经使用 opencv 在我的程序中加载预先存在的 2D 图像。我使用开放的 cl 缓冲区对象来分配内存并将图像指针发送到内核。在内核执行之后，为了显示来自内核的计算图像，我需要 clEnqueueReadBuffer。然后我使用 opencv 显示来自主机的图像。我在下面附上了代码

由于这需要更多时间在 GPU 和 CPU 上，我想切换到图像内存。

但我想知道图像的使用是否还需要 clenqueueReadImage 将图像从内核复制到主机，或者我们有什么方法可以在内核本身中显示阈值图像？

//My code using opencl buffers    
IplImage *src = cvLoadImage("../Input/im2.png",CV_LOAD_IMAGE_COLOR );

int a=src->height;
int b=src->width;

cl_context CreateContext()
{
    cl_int errNum;
    cl_uint numPlatforms;
    cl_platform_id firstPlatformId;
    cl_context context = NULL;
    errNum = clGetPlatformIDs(1, &firstPlatformId, &numPlatforms);
    if (errNum != CL_SUCCESS || numPlatforms <= 0)
    {
        std::cerr << "Failed to find any OpenCL platforms." << std::endl;
        return NULL;
    }

    cl_context_properties contextProperties[] =
    {
        CL_CONTEXT_PLATFORM,
        (cl_context_properties)firstPlatformId,
        0
    };
    context = clCreateContextFromType(contextProperties, CL_DEVICE_TYPE_GPU,
                                      NULL, NULL, &errNum);
    if (errNum != CL_SUCCESS)
    {
        std::cout << "Could not create GPU context, trying CPU..." << std::endl;
        context = clCreateContextFromType(contextProperties, CL_DEVICE_TYPE_CPU, NULL, NULL, &errNum);
        if (errNum != CL_SUCCESS)
        {
            std::cerr << "Failed to create an OpenCL GPU or CPU context." << std::endl;
            return NULL;
        }
    }
    return context;
}


cl_command_queue CreateCommandQueue(cl_context context, cl_device_id *device)
{
    cl_int errNum;
    cl_device_id *devices;
    cl_command_queue commandQueue = NULL;
    size_t deviceBufferSize = -1;
    errNum = clGetContextInfo(context, CL_CONTEXT_DEVICES, 0, NULL, &deviceBufferSize);
    if (errNum != CL_SUCCESS)
    {
        std::cerr << "Failed call to clGetContextInfo(...,GL_CONTEXT_DEVICES,...)";
        return NULL;
    }

    if (deviceBufferSize <= 0)
    {
        std::cerr << "No devices available.";
        return NULL;
    }
    devices = new cl_device_id[deviceBufferSize / sizeof(cl_device_id)];
    errNum = clGetContextInfo(context, CL_CONTEXT_DEVICES, deviceBufferSize, devices, NULL);
    if (errNum != CL_SUCCESS)
    {
        delete [] devices;
        std::cerr << "Failed to get device IDs";
        return NULL;
    }

    commandQueue = clCreateCommandQueue(context, devices[0],CL_QUEUE_PROFILING_ENABLE, &errNum );

    if (commandQueue == NULL)
    {
        delete [] devices;
        std::cerr << "Failed to create commandQueue for device 0";
        return NULL;
    }

    *device = devices[0];
    delete [] devices;
    return commandQueue;
}

cl_program CreateProgram(cl_context context, cl_device_id device, const char* fileName)
{
    cl_int errNum;
    cl_program program;

    std::ifstream kernelFile(fileName, std::ios::in);
    if (!kernelFile.is_open())
    {
        std::cerr << "Failed to open file for reading: " << fileName << std::endl;
        return NULL;
    }

    std::ostringstream oss;
    oss << kernelFile.rdbuf();

    std::string srcStdStr = oss.str();
    const char *srcStr = srcStdStr.c_str();
    program = clCreateProgramWithSource(context, 1,
                                        (const char**)&srcStr,
                                        NULL, NULL);
    if (program == NULL)
    {
        std::cerr << "Failed to create CL program from source." << std::endl;
        return NULL;
    }
    errNum = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);

    if (errNum != CL_SUCCESS)
    {
        char buildLog[16384];
        clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_LOG,
                              sizeof(buildLog), buildLog, NULL);
        std::cerr << "Error in kernel: " << std::endl;
        std::cerr << buildLog;
        clReleaseProgram(program);
        return NULL;
    }
    return program;
}

bool CreateMemObjects(cl_context context, cl_mem memObjects[2], unsigned char *src_ptr)
{
    memObjects[0] = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR, sizeof(unsigned char) *(a*b*3) , src_ptr , NULL);
    memObjects[1] = clCreateBuffer(context, CL_MEM_READ_WRITE, sizeof(unsigned char) *(a*b) , NULL, NULL);

   if (memObjects[0] == NULL || memObjects[1] == NULL)
    {
        std::cerr << "Error creating memory objects" << std::endl;
        return false;
    }
    return true;
}

void Cleanup(cl_context context, cl_command_queue commandQueue, cl_program program, cl_kernel kernel, cl_mem memObjects[2])
{
    for (int i = 0; i < 2; i++)
    {
        if (memObjects[i] != 0)
            clReleaseMemObject(memObjects[i]);
    }
    if (commandQueue != 0)
        clReleaseCommandQueue(commandQueue);

    if (kernel != 0)
        clReleaseKernel(kernel);

    if (program != 0)
        clReleaseProgram(program);

    if (context != 0)
        clReleaseContext(context);
}


int main()
{
    cl_context context = 0;
    cl_command_queue commandQueue = 0;
    cl_program program = 0;
    cl_device_id device = 0;
    cl_kernel kernel = 0;
    cl_mem memObjects[2] = { 0,0 };
    cl_int errNum;
    cl_event myEvent;

    cl_ulong start_time,end_time;
    double kernelExecTimeNs;

    IplImage *thres_img1 = cvCreateImage(cvGetSize(src), IPL_DEPTH_8U, 1);

    unsigned char *tur_image1,*src_ptr;
    tur_image1 = (unsigned char*) malloc((a*b) * sizeof(unsigned char));
    src_ptr = (unsigned char*) malloc ((a*b*3) * sizeof(unsigned char));

    context = CreateContext();
    if (context == NULL)
    {
        std::cerr << "Failed to create OpenCL context." <<std::endl;
        return 1;
    }

    commandQueue = CreateCommandQueue(context, &device);
    if (commandQueue == NULL)
    {
        Cleanup(context, commandQueue, program, kernel, memObjects);
        return 1;
    }

    program = CreateProgram(context, device, "hsl_threshold.cl");
    if (program == NULL)
    {
        Cleanup(context, commandQueue, program, kernel, memObjects);
        return 1;
    }

    kernel = clCreateKernel(program, "HSL_threshold", NULL);
    if (kernel == NULL)
    {
        std::cerr << "Failed to create kernel" << std::endl;
        Cleanup(context, commandQueue, program, kernel, memObjects);
        return 1;
    }

    printf("height:%d\n",a);//image height
    printf("width:%d\n",b);//image width

    cvShowImage("color image",src);
    cvWaitKey(0);

    memcpy(src_ptr,src->imageData,(a*b*3));

    if (!CreateMemObjects(context, memObjects, src_ptr))
    {
        Cleanup(context, commandQueue, program, kernel, memObjects);
        return 1;
    }

        errNum  = clSetKernelArg(kernel, 0, sizeof(cl_mem), &memObjects[0]);
    errNum |= clSetKernelArg(kernel, 1, sizeof(cl_mem), &memObjects[1]);

    if (errNum != CL_SUCCESS)
    {
        std::cerr << "Error setting kernel arguments" << std::endl;
        Cleanup(context, commandQueue, program, kernel, memObjects);
        return 1;
    }

    cout<<"Kernel arguments set successfully";
    size_t globalWorkSize[1]={a*b};
    size_t localWorkSize[1]={512};

    errNum = clEnqueueNDRangeKernel(commandQueue, kernel, 1, NULL, globalWorkSize, localWorkSize, 0, NULL, &myEvent);

    clWaitForEvents(1,&myEvent);

    if (errNum != CL_SUCCESS)
    {
        std::cerr << "Error queuing kernel for execution." << std::endl;
        Cleanup(context, commandQueue, program, kernel, memObjects);
        return 1;
    }
    clFinish(commandQueue);


    clGetEventProfilingInfo(myEvent, CL_PROFILING_COMMAND_START, sizeof(start_time), &start_time, NULL);
    clGetEventProfilingInfo(myEvent, CL_PROFILING_COMMAND_END, sizeof(end_time), &end_time, NULL);

    kernelExecTimeNs = end_time-start_time;

    printf("\nExecution time in milliseconds = %0.3f ms\n",( kernelExecTimeNs / 1000000.0) );
    cout<<"\n Kernel timings \n"<<kernelExecTimeNs<<"seconds";

    errNum = clEnqueueReadBuffer(commandQueue, memObjects[1], CL_TRUE,
                                 0, (a*b) * sizeof(unsigned char), tur_image1,
                                 0, NULL, NULL);

    if (errNum != CL_SUCCESS)
    {
        std::cerr << "Error reading result buffer." << std::endl;
        Cleanup(context, commandQueue, program, kernel, memObjects);
        return 1;
    }

    memcpy(thres_img1->imageData,tur_image1,sizeof(unsigned char)*(a*b));

    cvShowImage( "hsl_thresh",thres_img1);
    cvSaveImage( "../Output/hsl_threshold.png",thres_img1);
    cvWaitKey(0);

    std::cout<<std::endl;
    std::cout<<"Image displayed Successfully"<<std::endl;

    Cleanup(context,commandQueue,program,kernel,memObjects);
    printf("\n Free opencl resources");
    std::cin.get();
    return 0;

}

score 7 · Accepted Answer

有一些方法可以通过 OpenGL 直接处理 OpenCL 计算的数据。您的 OCL 实现必须支持扩展cl_khr_gl_sharing。
这种模式称为 CL/GL 互操作模式。

如果您首先创建一个 OpenGL 实例并使用指向您的 GL 实例的指针初始化 OpenCL，则每个实现都可以访问彼此的数据。

（所有片段均取自使用 CL-C++-Bindings 的代码，我想这对于一般理解来说是可以的）

cl_context_properties properties[] = 
  // Take this line to create an OCL context in GL-CL-interop-mode. 
  // OpenGL must already be initialised. 
  // For interop init see: http://www.khronos.org/registry/cl/extensions/khr/cl_khr_gl_sharing.txt
  // USING: CL_GL_CONTEXT_KHR: Rendering Context [Use your OGL-HGLRC variable or do wglGetCurrentContext(); ]
  //   AND: CL_WGL_HDC_KHR: Device Context [Use your OGL-HDC variable or do wglGetCurrentDC(); ]
  { 
    CL_CONTEXT_PLATFORM, (cl_context_properties)(_platforms->at(0))(), 
    CL_GL_CONTEXT_KHR, (cl_context_properties)myGL->hRC, 
    CL_WGL_HDC_KHR, (cl_context_properties)myGL->hDC, 0
  };

现在您可以基于 OGL 纹理创建 OCL 图像

//The following data can be accessed both from OCL and OGL
cl::Image2D imageFromGL = new cl::Image2DGL(*_context, CL_MEM_READ_WRITE,  GL_TEXTURE_2D, 0, myGL->textures[0]);

在 OCL 中使用内存之前，您必须要求 OGL 释放它

//Ask OGL to release memory. All OGL actions must be finished before doing so!
_queue->enqueueAcquireGLObjects(&imageFromGL, NULL, &evt);

现在，做你想做的，然后把它还给 OGL：

//Hand memory back to OGL. All OCL actions must be finished before doing so!
_queue->enqueueReleaseGLObjects(&imageFromGL, NULL, &evt);

最后，您可以使用 OpenGL 代码在屏幕上显示数据。

opencl - 如何使用 opencl 在内核中显示图像？

1 回答 1

Related

Reference