fastest way to set every pixel

Question

I have programmed a little raytracer in c++,
and want to show the raytraced image in a window.

I tried using a pixel buffer object in opengl,
then map the buffer into memory and manipulate the pixels one by one,
but at fullscreen resolution 1920x1080, I only get 4 fps
without raytracing and without changing the pixels colors
just the mapping and unmapping!

so i'm basically looking for the fastest way to display a raytraced image in a window.

i'm currently doing this way:

glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, pbo);
glBufferDataARB(GL_PIXEL_UNPACK_BUFFER_ARB, width * height * 4, 0, GL_STREAM_DRAW_ARB);
if (pixels = (uint*)glMapBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, GL_WRITE_ONLY_ARB))
{
    //modify pixels
    glUnmapBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB);
}
else
    return;


//copy from pbo to texture
glBindTexture(GL_TEXTURE_2D, pbo_texture);
glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, pbo);
glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, width, height, GL_BGRA, GL_UNSIGNED_BYTE, 0);

glEnable(GL_TEXTURE_2D);

glMatrixMode(GL_PROJECTION);
glLoadIdentity();

glMatrixMode(GL_MODELVIEW);
glLoadIdentity();

//draw image
glColor4f(1.0, 1.0, 1.0, 1.0);
glBindTexture(GL_TEXTURE_2D, pbo_texture);
glBegin(GL_QUADS);
    glTexCoord2f(0.0, 1.0); glVertex3f(-1.0, -1.0, 0.0);
    glTexCoord2f(1.0, 1.0); glVertex3f( 1.0, -1.0, 0.0);
    glTexCoord2f(1.0, 0.0); glVertex3f( 1.0,  1.0, 0.0);
    glTexCoord2f(0.0, 0.0); glVertex3f(-1.0,  1.0, 0.0);
glEnd();

glutSwapBuffers();

glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, 0);
glBindTexture(GL_TEXTURE_2D, 0);

Answer 1

Check the memory traversal if you use loops. You should traverse your buffer in the right order, otherwise you may have cache miss at each iteration. If you use nested loops sometimes you only have to switch the x/y iteration order.

Also, don't read data from graphic memory. It tends to be slow. Only write to PBO.

It looks like a syncro issue. I'm not sure you need to map pbo at every frame. Check this link on OpenGL Pixel Buffer Object (PBO). There's also a workaround for stalls which could improve things