I-Corps: Hardware Acceleration of Scientific Computing and Intelligent Instrumentation

Award ID: 2323077

The broader impact/commercial potential of this I-Corps project is the development of faster microprocessors for scientific computing applications. Currently, scientific computing relies heavily on supercomputers comprised of large clusters of general-purpose microprocessors. Such an approach is expensive, and does not deliver the optimal performance for scientific computing applications. In the proposed technology, the hardware-software framework under development may provide a faster way for physicists, biologists, chemists, environmental scientists, and interdisciplinary researchers to run their simulations faster, and with greater ease. In addition, the proposed technology may provide for more enriched computing systems integrated with multiple types of microprocessors, not simply general-purpose microprocessors or graphics-focused graphics processing units. Specialized microprocessor designs may be suitable for low-power applications in portable instrumentation devices. It is anticipated that customized chips to be integrated into a broad array of application areas such as aerospace, pharmaceutical and multi-physics systems with a need for real-time and predictive control. This I-Corps project is based on the development of application-specific microprocessors including a robust programming framework. The proposed technology addresses the development of relevant algorithms to analyze the computational behavior of targeted software applications. The application-specific microprocessors developed for this project will focus primarily on accelerating the numerical solution of differential equations. Differential equations form the mathematical basis for many computational dynamics systems, and it is important to solve them in a rapid and efficient manner. The popularization of open-source instruction set architectures has enabled new microprocessor design while the advancement of field programmable gate array (FPGA) technology has enabled rapid prototyping and evaluation. The intersection of these developments has facilitated development of target application-specific microprocessor designs at low cost. Initial results have shown that the designed hardware accelerator attains up to 4.8x speedup, at the cost of merely 13.3% more hardware resources and 8.1% additional power dissipation. In the future, it also may be possible to design low-power microprocessors that can perform signal analysis on-board measurement instruments. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Program Director: Ruth Shuman