Architectural Acceleration of Real-Time Physics

Physics-based animation (PBA) is becoming one of the most important elements of interactive entertainment applications, largely because of the automation and realism that it offers. The immersive nature of interactive entertainment relies on rapid, dynamic content creation and realistic visual effects. However, the benefits of PBA come at a
considerable computational cost, and the demands of interactive entertainment require soft performance bounds of 30-60 frames per second.

We have created a forward-thinking benchmarking suite for PBA to characterize this emerging workload, and identify two main characteristics that we can exploit at the architectural level. First, PBA has massive amounts of parallelism, but this parallelism must be carefully leveraged as there is some serialization that must occur.
Second, PBA effects need to be believable, but not 100% accurate. However, the trick is determining how much accuracy is needed (and where) for believability, and how to capitalize on this architecturally.

In this talk, I will detail the ParallAX architecture, a chip multiprocessor design targeted towards PBA. This design exploits both the massive parallelism and error tolerance of PBA. I will also discuss some of the other architectural research being conducted at UCLA, including RF interconnect shortcuts for chip multiprocessors.

Bio:
Glenn Reinman is an assistant professor in the Department of Computer
Science at University of California, Los Angeles. He received his Ph.D.
and M.S. in Computer Science at the University of California, San Diego
in 2001 with a dissertation on fetch optimizations for aggressive
out-of-order superscalar processors. He received a B.S. in Computer
Science and Engineering from the Massachusetts Institute of Technology
in 1996.

Glenn was awarded an NSF CAREER grant in 2001 titled The Evaluation and
Design of a Scalable, High-Performance, and Energy-Efficient
Microprocessor Architecture. He has also received the Northrop Grumman
Excellence in Teaching Award in 2004 and was named Professor of the Year
by the undergraduate chapter of the Engineering Society of the
University of California Los Angeles in 2006.