A combination of parallelism exploitation and application specific hardware is increasingly being used to address the computational requirements of a diverse and extensive set of application areas. These targeted applications have specific computational requirements that often are not able to be implemented optimally on general purpose processors and have the potential to experience substantial speedup on dedicated hardware. While general parallelism has been exploited at various levels for decades, the advent of heterogeneous cluster computing has allowed applications to be accelerated through the use of intelligently mapped computational tasks to well-suited hardware. This trend has continued with the use of dedicated ASIC and FPGA coprocessors to off-load particularly intensive computations. With the inclusion of embedded microprocessors into otherwise reconfigurable FPGA fabric, it has become feasible to construct a heterogeneous cluster composed of application specific hardware resources that can be programatically treated as fully functional and independent cluster nodes via a standard message passing interface. The contribution of this thesis is the development of such a framework for organizing heterogeneous clusters of reconfigurable FPGA computing elements into clusters that enable development of complex systems delivering on the promise of parallel reconfigurable hardware. The framework includes a fully featured message passing interface implementation for seamless communication and synchronization among nodes running in an embedded Linux operating system environment while managing hardware accelerators through device driver abstractions and standard APIs. A set of application case studies deployed on a test platform of Xilinx Virtex-4 and Virtex-5 FPGAs demonstrates functionality, elucidates performance characteristics, and promotes future research and development efforts.
Library of Congress Subject Headings
Parallel processing (Electronic computers); Field programmable gate arrays; Computer networks
Department, Program, or Center
Computer Engineering (KGCOE)
Espenshade, Jeremy K., "Scalable framework for heterogeneous clustering of commodity FPGAs" (2009). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus