Ubuntu virtualbox中的OpenGL多线程示例

Question

I am facing a weird problem with an OpenGL sample on vbox with 3D acceleration enabled

• Guest: Ubuntu 12.04
• Host: Windows 7, nvidia Graphics
• vbox version 4.3.6 with guest additions installed

This application runs properly on vbox when 3D acceleration is disabled and I have checked this on stand alone Linux PC as well. When the same is run with 3D acceleration enabled, its not able to get the GL function pointers giving errors - function no-op

---

The App is Simple, Main thread creates 2 threads.

• Main Thread - Create Thread 1, Create Thread 2
• Thread 1 - Create a X-Window for Rendering
• Thread 2 - Create Render thread (Draw OpenGL quad on the X-Window).

Here is the code for the sample app.

#include<stdio.h>
#include<stdlib.h>
#include<X11/X.h>
#include<X11/Xlib.h>
#include<GL/gl.h>
#include<GL/glx.h>
//#include<GL/glu.h>
#include <dlfcn.h> /*dlopen*/
#include <pthread.h>
#include <unistd.h> /*sleep*/

Display                 *dpy;
Display                 *dpy2;
Window                  root;
GLint                   att[] = { GLX_RGBA, GLX_DEPTH_SIZE, 24, GLX_DOUBLEBUFFER, None };
XVisualInfo             *vi;
XVisualInfo             *vi2;
Colormap                cmap;
XSetWindowAttributes    swa;
Window                  win;
GLXContext              glc;
XWindowAttributes       gwa;
XEvent                  xev;
bool            render;


void DrawAQuad() 
{
    glClearColor(1.0, 1.0, 1.0, 1.0);
    glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);

    glMatrixMode(GL_PROJECTION);
    glLoadIdentity();
    glOrtho(-1., 1., -1., 1., 1., 20.);

    glMatrixMode(GL_MODELVIEW);
    glLoadIdentity();
    //gluLookAt(0., 0., 10., 0., 0., 0., 0., 1., 0.);
    glTranslatef(0.0, 0.0, -10.0);

    glBegin(GL_QUADS);
     glColor3f(1., 0., 0.); glVertex3f(-.75, -.75, 0.);
     glColor3f(0., 1., 0.); glVertex3f( .75, -.75, 0.);
     glColor3f(0., 0., 1.); glVertex3f( .75,  .75, 0.);
     glColor3f(1., 1., 0.); glVertex3f(-.75,  .75, 0.);
    glEnd();
}

void *CreateMainWindow(void* threadID)
{
    dpy = XOpenDisplay(NULL);
    if(dpy == NULL) 
    {
        printf("\n\tWindow Thread: cannot connect to X server\n\n");
        exit(0);
    }

    root = DefaultRootWindow(dpy);
    printf("\n *** CreateWindow: xopendisplay over *** \n");

    vi = (XVisualInfo*)glXChooseVisual(dpy, 0, att);
    if(vi == NULL) 
    {
        printf("\n\tWindow Thread: no appropriate visual found\n\n");
        exit(0);
    } 
    else 
    {
        printf("\n\tWindow Thread: visual %p selected\n", (void *)vi->visualid);
    }

    cmap = XCreateColormap(dpy, root, vi->visual, AllocNone);
    swa.colormap = cmap;
    swa.event_mask = ExposureMask | KeyPressMask;

    win = XCreateWindow(dpy, root, 0, 0, 600, 600, 0, vi->depth, InputOutput, vi->visual, CWColormap | CWEventMask, &swa);
    XMapWindow(dpy, win);
    XStoreName(dpy, win, "VERY SIMPLE APPLICATION");

    while(1) 
    {
        XNextEvent(dpy, &xev);
        printf("\nXEVENT\n");
        if(xev.type == Expose)
        {
            render = true;
        }
        else if(xev.type == KeyPress)
        {
            XDestroyWindow(dpy, win);
            XCloseDisplay(dpy);
            render = false;
            break;
            //exit(0);
        }
    }
}

void *RenderThread(void* threadID)
{
    vi2 = (XVisualInfo*)glXChooseVisual(dpy2, 0, att);
    printf("\n\tRenderThread : visual %p selected\n", (void *)vi2->visualid);

    glc = (GLXContext)glXCreateContext(dpy2, vi2, NULL, GL_TRUE);
    glXMakeCurrent(dpy2, win, glc);

    glEnable(GL_DEPTH_TEST); 

    while(render) 
    {
        //XGetWindowAttributes(dpy, win, &gwa);
        glViewport(0, 0, 600, 600);
        DrawAQuad(); 
        glXSwapBuffers(dpy2, win);
    } /* this closes while(render) */

    glXMakeCurrent(dpy2, None, NULL);
    glXDestroyContext(dpy2, glc);
    XCloseDisplay(dpy2);
}

int main(int argc, char *argv[]) 
{
    render = true;
    pthread_t thread1;
    pthread_t thread2;
    char *temp1;
    char *temp2;

    //For Async issue
    if(!XInitThreads())
    {
        fprintf(stderr, "XInitThread failed\n");
        return 0;
    }

    //Create Main Window
    int err = pthread_create(&thread1, NULL, CreateMainWindow, (void*)temp1);
    if (err != 0)
        printf("\n ERROR::can't create thread1 :[%d]", err);
    else
        printf("\n Thread1 created successfully\n");

    sleep(1); // Wait for thread 1 to complete

    dpy2 = XOpenDisplay(NULL);
    if(dpy2 == NULL) 
    {
        printf("\n\tMain : cannot connect to X server\n\n");
        exit(0);
    }

    //Create Render Thread
    err = pthread_create(&thread2, NULL, RenderThread, (void*)temp2);
    if (err != 0)
        printf("\n ERROR::can't create thread2 :[%d]", err);
    else
        printf("\n Thread2 created successfully\n");

    pthread_join( thread1, NULL);
    pthread_join( thread2, NULL);

} /* this is the } which closes int main(int argc, char *argv[]) { */

and to compile the code -

g++ -o quad quad.cpp -lGL -lX11 -lXmu -lXi -lpthread -lm

Help me understand where the issue lies

Answer 1

Whatever you thick you're doing, STOP!

Multithreading + X11 + OpenGL is a very tricky thing to get right. And it's nearly impossible to do correct if you're using Xlib. Xlib never was truly thread safe.

Anyway, first and foremost your program lacks a call to XInitThreads to make it at least safe to use in multithreaded programs. It's however still unsafe to spread out Xlib calls over multiple threads. This is really important: Whatever you do, keep all Xlib calls to one thread only.

OpenGL itself is not as tricky. But because OpenGL needs a context created with glX, which in turn builds on Xlib. The usual approach is to create the OpenGL context in the Xlib thread, but later on make it current in the renderer thread. However be advised that if you got an indirect rendering context all OpenGL calls go through X11 and that may mean through Xlib and things are going to be unstable again.

Because of all that mess the simplest solution is: Keep everything graphics and windowing related to one thread. There's nothing to be gained if you put OpenGL operations into a thread separate from the rest of the GUI operations (technically OpenGL does GUI operations as well). If you want to use multithreading, then use if for thing that make sense to be executed concurrently, like audio, physics simulation and such.