The Siggraph 2003 presentation “Multipass Rendering in mental ray for “The Matrix” sequels”, an early deep frame buffer approach to rendering and compositing large datasets, are online here:

Authors:

Abstract:
Rendering resources are by definition constrained and yet the task of rendering many objects at high quality is frequently encountered in the effects industry. Breaking up scenes into renderable pieces, or passes, is a common approach to rendering large scenes. To achieve the quantity of photo-real effects in frame that were required by the scripts of The Matrix Reloaded and The Matrix Revolutions, mental images and ESC Entertainment developed a system for rendering and automatically compositing many passes. The system we developed has the key advantage that it takes as its compositing primitive not the pixel but the sample.

The full writeup and slides are available below
Continue reading →