Embedded intelligence applications must optimize the energy efficiency of their computational differentiation. IEEE-754 floating-point has been the workhorse for robust numerical computing, but is notoriously energy inefficient. Next-generation arithmetic solutions, such as posits and cfloats combined with user-defined rounding have demonstrated a 2-4x power benefit over IEEE-754. We have developed a mixed-precision algorithm design and optimization environment to deliver this technology to the marketplace in the form of custom applications and hardware accelerators. We'll use this platform for key SmartCity application such as collective intelligence and congestion optimization.

Project Proposal

1. High-level project introduction and performance expectation

2. Block Diagram

3. Expected sustainability results, projected resource savings