Fastest way to manipulate pixels with cuda + opengl + driver api ?

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello,

I am currently wondering what would be the fastest way to manipulate pixels with cuda + opengl + driver api (for cuda >= 4.0 and any opengl+any extensions) ?

So far I have seen some possible ideas:

  1. Texture + Draw texture to screen
  2. Texture and/or Render buffer + Framebuffer object
  3. Pixel buffer object + draw
  4. Surface (?)
  5. Buffer object (?)

Bye,
Skybuck.