This video was recorded using the system described in the paper.
  Its content shows the camera path used to measure the efficiency of the partial composition,
of the final image splitting optimization and of the system stress test (sections 5.2,
5.3 and 5.4, respectively).
  The viewed model consists on approximately 42 million active cells.
  At the right, it is possible to observe some debugging info, such as the time being spent
on some pipeline stages and each GPU rendering time.

  The zoom into the reservoir shows the big amount of primitives being rendered at each frame.

  Observation: the video frame rate should not be considered, since the video recording is
degrading the overall performance.