像素缓冲区对象（PBO）的异步Read-Back 源码解析

接这篇文章 OpenGL深入探索——像素缓冲区对象 （PBO）（附完整工程代码地址）

原理示意图如下：

关键代码如下：

int main(int argc, char **argv)
{
    initSharedMem();

    // register exit callback
    atexit(exitCB);

    // init GLUT and GL
    initGLUT(argc, argv);
    initGL();

    // get OpenGL info
    glInfo glInfo;
    glInfo.getInfo();
    glInfo.printSelf();

#ifdef _WIN32
    // 检查视频显卡是否支持 PBO
    if (glInfo.isExtensionSupported("GL_ARB_pixel_buffer_object"))
    {
        // get pointers to GL functions
        glGenBuffersARB = (PFNGLGENBUFFERSARBPROC)wglGetProcAddress("glGenBuffersARB");
        glBindBufferARB = (PFNGLBINDBUFFERARBPROC)wglGetProcAddress("glBindBufferARB");
        glBufferDataARB = (PFNGLBUFFERDATAARBPROC)wglGetProcAddress("glBufferDataARB");
        glBufferSubDataARB = (PFNGLBUFFERSUBDATAARBPROC)wglGetProcAddress("glBufferSubDataARB");
        glDeleteBuffersARB = (PFNGLDELETEBUFFERSARBPROC)wglGetProcAddress("glDeleteBuffersARB");
        glGetBufferParameterivARB = (PFNGLGETBUFFERPARAMETERIVARBPROC)wglGetProcAddress("glGetBufferParameterivARB");
        glMapBufferARB = (PFNGLMAPBUFFERARBPROC)wglGetProcAddress("glMapBufferARB");
        glUnmapBufferARB = (PFNGLUNMAPBUFFERARBPROC)wglGetProcAddress("glUnmapBufferARB");

        // check once again PBO extension
        if (glGenBuffersARB && glBindBufferARB && glBufferDataARB && glBufferSubDataARB &&
                glMapBufferARB && glUnmapBufferARB && glDeleteBuffersARB && glGetBufferParameterivARB)
        {
            pboSupported = pboUsed = true;
            std::cout << "Video card supports GL_ARB_pixel_buffer_object." << std::endl;
        }
        else
        {
            pboSupported = pboUsed = false;
            std::cout << "Video card does NOT support GL_ARB_pixel_buffer_object." << std::endl;
        }
    }

    // check EXT_swap_control is supported
    if (glInfo.isExtensionSupported("WGL_EXT_swap_control"))
    {
        // get pointers to WGL functions
        wglSwapIntervalEXT = (PFNWGLSWAPINTERVALEXTPROC)wglGetProcAddress("wglSwapIntervalEXT");
        wglGetSwapIntervalEXT = (PFNWGLGETSWAPINTERVALEXTPROC)wglGetProcAddress("wglGetSwapIntervalEXT");
        if (wglSwapIntervalEXT && wglGetSwapIntervalEXT)
        {
            // disable v-sync
            wglSwapIntervalEXT(0);
            std::cout << "Video card supports WGL_EXT_swap_control." << std::endl;
        }
    }

#else // for linux, do not need to get function pointers, it is up-to-date
    if (glInfo.isExtensionSupported("GL_ARB_pixel_buffer_object"))
    {
        pboSupported = pboUsed = true;
        std::cout << "Video card supports GL_ARB_pixel_buffer_object." << std::endl;
    }
    else
    {
        pboSupported = pboUsed = false;
        std::cout << "Video card does NOT support GL_ARB_pixel_buffer_object." << std::endl;
    }
#endif

    if (pboSupported)
    {
        // create 2 pixel buffer objects, you need to delete them when program exits.
		//（创建两个 PBO）
        // glBufferDataARB with NULL pointer reserves only memory space.
        glGenBuffersARB(PBO_COUNT, pboIds);
        glBindBufferARB(GL_PIXEL_PACK_BUFFER_ARB, pboIds[0]);
        glBufferDataARB(GL_PIXEL_PACK_BUFFER_ARB, DATA_SIZE, 0, GL_STREAM_READ_ARB);
        glBindBufferARB(GL_PIXEL_PACK_BUFFER_ARB, pboIds[1]);
        glBufferDataARB(GL_PIXEL_PACK_BUFFER_ARB, DATA_SIZE, 0, GL_STREAM_READ_ARB);

        glBindBufferARB(GL_PIXEL_PACK_BUFFER_ARB, 0);
    }

    // start timer, the elapsed time will be used for updateVertices()
    timer.start();

    // the last GLUT call (LOOP)
    // window will be shown and display callback is triggered by events
    // NOTE: this call never return main().
    glutMainLoop(); /* Start GLUT event-processing loop */

    return 0;
}

省略掉中间其他一些无关代码，剩下关键方法：

void displayCB()
{
    static int shift = 0;
    static int index = 0;
    int nextIndex = 0;                  // pbo index used for next frame

    // brightness shift amount
    shift = ++shift % 200;

    // increment current index first then get the next index（增加当前的index，再获得 nextIndex）
    // "index" is used to read pixels from a framebuffer to a PBO（从 FB 中读取像素到 index 指定的 PBO 中）
    // "nextIndex" is used to process pixels in the other PBO（nextIndex 指定 PBO 中待处理的像素[先前从 FB 中读取]）
    index = (index + 1) % 2;	// 两个 PBO 交替
    nextIndex = (index + 1) % 2;

    // set the framebuffer to read（设置当前读取的 FB）
    glReadBuffer(GL_FRONT);

    if (pboUsed) // with PBO
    {
        // read framebuffer ///
        t1.start();  // 启动计时器，计算读取 FB 到 PBO 的时间

        // copy pixels from framebuffer to PBO （从 FB 中拷贝像素到 index指定的 PBO 中）
        // Use offset instead of ponter.（注意glReadPixels最后一个参数不是指针，而是偏移量）
        // OpenGL should perform asynch DMA transfer, so glReadPixels() will return immediately.
		//（OpenGL 将执行一个异步的DMA[Direcct Memory Access]，所以 glReadPixels方法会立刻返回，不会阻塞CPU时间）
        glBindBufferARB(GL_PIXEL_PACK_BUFFER_ARB, pboIds[index]);
        glReadPixels(0, 0, SCREEN_WIDTH, SCREEN_HEIGHT, PIXEL_FORMAT, GL_UNSIGNED_BYTE, 0);

        // measure the time reading framebuffer
        t1.stop();
        readTime = t1.getElapsedTimeInMilliSec();
        ///

        // process pixel data /
        t1.start(); // 启动计时器，计算处理 PBO中像素的时间

        // map the PBO that contain framebuffer pixels before processing it
		//（映射存储着先前 FB 像素的 PBO ，便于处理像素）
        glBindBufferARB(GL_PIXEL_PACK_BUFFER_ARB, pboIds[nextIndex]);
		// 后续的 add() 并没有改变 PBO 中的像素值，计算的结果保存在 colorBuffer 中，所以标记为只读
        GLubyte *src = (GLubyte *)glMapBufferARB(GL_PIXEL_PACK_BUFFER_ARB, GL_READ_ONLY_ARB);
        if (src)
        {
            // change brightness
            add(src, SCREEN_WIDTH, SCREEN_HEIGHT, shift, colorBuffer);
            glUnmapBufferARB(GL_PIXEL_PACK_BUFFER_ARB);     // release pointer to the mapped buffer
        }

        // measure the time processing the pixels of PBO
        t1.stop();
        processTime = t1.getElapsedTimeInMilliSec();
        ///

        glBindBufferARB(GL_PIXEL_PACK_BUFFER_ARB, 0);
    }
    else        // without PBO
    {
        // read framebuffer ///
        t1.start();
		// 不使用 PBO 的情况下，最后一个参数就是保存读取数据的指针
        glReadPixels(0, 0, SCREEN_WIDTH, SCREEN_HEIGHT, PIXEL_FORMAT, GL_UNSIGNED_BYTE, colorBuffer);

        // measure the time reading framebuffer
        t1.stop();
        readTime = t1.getElapsedTimeInMilliSec();
        ///

        // covert to greyscale 
        t1.start();

        // change brightness
        add(colorBuffer, SCREEN_WIDTH, SCREEN_HEIGHT, shift, colorBuffer);

        // measure the time reading framebuffer
        t1.stop();
        processTime = t1.getElapsedTimeInMilliSec();
        ///
    }

    // render to the framebuffer //
    glDrawBuffer(GL_BACK);	// 设定绘制在后缓冲区
    toPerspective(); // set to perspective on the left side of the window（当前窗口左侧的投影矩阵）

    // clear buffer
    glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT);

    // tramsform camera
    glTranslatef(0, 0, -cameraDistance);
    glRotatef(cameraAngleX, 1, 0, 0);   // pitch
    glRotatef(cameraAngleY, 0, 1, 0);   // heading

    // draw a cube（在左侧绘制普通的立方体）
    glPushMatrix();
    draw();
    glPopMatrix();

    // draw the read color buffer to the right side of the window
	//（在右侧绘制之前处理的 colorBuffer）
    toOrtho();      // set to orthographic on the right side of the window（窗口右侧的正交矩阵[想象为把左侧的图处理一下直接贴上去]）
    glRasterPos2i(0, 0);	// 设置字体光栅的位置
    glDrawPixels(SCREEN_WIDTH, SCREEN_HEIGHT, PIXEL_FORMAT, GL_UNSIGNED_BYTE, colorBuffer);

    // draw info messages
    showInfo();
    printTransferRate();

    glutSwapBuffers();// 切换前后缓冲区
}

程序的执行对比情况：

 可见通过 PBO的异步read-back技术，FPS 还是有明显提高的（我的显卡是 GeForce GTX 970，CPU 是 i7-6700 HQ，主频为2.6GHz）。