Our research group is pursuing a "vertically integrated" approach to
real-time 3-D graphics systems research. Our work spans graphics algorithms, software systems, and hardware architectures, as well as the interaction between these system layers.
Research summary
We are exploring new system designs for real-time graphics that:
Allow the use of visibility algorithms that are better than the
traditional Z-buffer.
Examples of this approach include our irregular Z buffer work, and various forms of raytracing that we are exploring.
Efficiently support the creation, modification, and traversal of
adaptive data structures such as linked lists and kd-trees.
Efficiently support a wide variety of parallel algorithms.
Our architectural focus is on single-chip parallel architectures. Our graphics focus is on complex, real-time scenes containing moving and deformable objects. Almost all of our projects require innovations in both graphics algorithms and hardware architectures.
Longer-term goal
The needs of real-time 3-D graphics are likely to drive the future
evolution of commodity parallel architectures. Fortunately, we anticipate
that the architectural capabilities needed to support our real-time
graphics goals will be sufficiently general that future graphics-oriented
architectures will be useful for a broad variety of other parallelizable
computations. Thus, we believe that we are in effect designing the commodity
general-purpose parallel computer architecture of the future.