I don't have a framework in mind. If I were to do this, I would write raw cuda with opengl introp. the output of cuda will be in a texture map memory and given

I don't have a framework in mind. If I were to do this, I would write raw cuda with opengl introp. the output of cuda will be in a texture map memory and given to opengl for rendering. although I have not done anything with visualization using this approach, I have, however, did video processing with this approach. A video frame is first processed by cuda and then passed to opengl for rendering.