The proposed NASA SBIR Phase-2 project will advance the state-of-the-art of computing acceleration to overcome the barriers of the end of Moore’s Law and power bounded clock rates through the invention of a non-von Neumann architecture that eliminates bottlenecks of incremental practices of conventional architecture. The family of Continuum Computer Architecture (CCA) exposes two orders magnitude efficiency, scalability, and user productivity with respect to energy, space, cost, and user productivity. CCA leapfrogs competing accelerators for irregular time-varying graph processing, their changing metadata through message-driven computing directed by graph-structure metadata and exploiting its intrinsic data parallelism for greatly increased scalability. This project will conduct technology engineering development from early analysis, low-level software, and preliminary breadboarding of parts equivalent accomplishments of Phase-1 to a medium level prototype incorporating all custom on-chip types of mechanisms and PC board closely related to the anticipated commercialization and marketing product to be delivered in Phase-3. The Phase-1 software environment for user programming and control will be enhanced for high language bindings and joint runtime system software between host control and accelerator adaptive task scheduling. This project will focus on NASA mission-critical ground-based, flight-, and space-borne applications including mechanisms for fault tolerance through graceful degradation and for real-time task scheduling. Other application domains will address those of the Intelligence community through its strong graph computing capabilities, GF2 operations and other related ops such as statistical histogramming and industry engineering challenges such as computational fluid dynamics, AMR, autonomous robotic control, and autonomous control. This project will position Simultac LLC to deliver its first product offerings as OEM and end user plug-n-play.
CCA-AMA is targeted to key and broad applications discussed in-depth with NASA AMES
-Data analytics, machine learning, and AI for ground-based HPC systems
-Planning and scheduling of spaceborne missions and real-time robotics
-Ground-based aeronautical material and computation fluid dynamics
-Neural net emulation
-Swarming for satellite constellations and free-flight aero-drones
-Gateway autonomous (when unmanned) condition monitoring and failure recovery
-Deep-space exploration of Titan; reconnaissance helicopter in Titan low atmosphere
-Embedded desk-side workstations or attached scalable mainframe graph acceleration
-Machine/Deep learning for expanding field of data-analytics including pervasive health
-Crash simulation of vehicles for automobile design cycles at lower cost
-Dynamic 3-D computational fluid dynamics and material bending for aeroplane CAD
-Financial markets and actuarial science for insurance