Overview

The center's research agenda is guided by three initial technical vectors, whose intersections will help realize non-conventional architectures that address these pressing challenges.

(i) Data-Centric Architectures: The current compute-centric architectures were established when data was small and compute was expensive. A reversal of this context demands a complete rethinking of architectures to focus on data from acquisition to result delivery. This rethinking has the potential to deliver orders-of-magnitude energy and performance efficiencies.

(ii) Novel Architectures Based on Emerging Technologies: Future architectures must maximally exploit the headroom offered by novel logic, interconnect (including 3D), and memory technologies, which provide non-volatility, density, power and performance advantages.

(iii) Beyond Homogeneous Parallelism: The benefits of homogenous parallelism are already being heavily mined through current research investments in parallel computing, leaving the next challenge of Amdahl's speedup constraints that must be addressed by reducing computation, communication and storage latencies.

Research Focus

The center research will target critical breakthroughs in all dimensions of future architectures, while maintaining efforts to ensure center designs are realizable and relevant, i.e., have implementations that meet design-time assumptions, function as expected, and address projected application needs. Finally, given significant design and deployment costs for applications and platforms, the solutions must amortize this cost over generations, over a range of capabilities in a generation, and with a manageable migration path from current systems. These high-level research needs are reflected in the research organization of the center. Research tasks within the center are housed within six research themes.

1. Computation
2. Communication
3. Storage
4. Applications to Architectures
5. Robust Design
6. System Integration