Extending the Thread Execution Model for Hybrid CPU/FPGA Architectures

University Of Kansas Center For Research Inc, Lawrence KS

Investigators

Abstract

Andrews Recently emerging CPU/FPGA hybrid chips are becoming important components of COTS hardware platforms for future embedded systems. Coupled with wireless networking capabilities, these hybrid chips offer significant computational advantages for a wide domain of distributed systems such as sensor webs. System software for these hybrids must provide a fairly general set of capabilities to support domain applications, but must also be capable of specialized support for each application's interfacing and performance needs. Creating such a capability for these applications is a difficult challenge in part, because it requires the simultaneous satisfaction of apparently contradictory forces: generalization and specialization. This research is developing new system software capabilities that promote the resolution of these seemingly contradictory requirements. Co-design of KU Real-Time (KURT) Linux system software components for software/FPGA implementation is studied to facilitate a familiar and portable multi-threaded programming model across the FPGA/CPU component boundaries. The research is developing synchronization and control methods for threads that must execute on both CPU and FPGA hardware. It focuses on enhanced scheduling capabilities and a programming environment for embedded sensor web applications that presents a familiar programming model for the new hardware architectures. This work is expected to enable more capable real-time control (on a chip) of sensing, computation, and data handling services than current widely-used sensor research platforms can support. In addition to generating new concepts and approaches to sensor web design for the research community, this effort supports an educational goal of establishing a focus area in the engineering of computer based systems. Within this focus area, we courses in embedded and real time systems including sensor webs, run time software for embedded systems, and distributed systems, are offered. This project broadens educational impact for both graduate and undergraduate students by integrating research activities and results into the classroom and producing interdisciplinary collaborative. The software tools, systems software, and hardware platforms developed under this project will be freely distributed to other universities and student laboratories for experimentation and projects.