Field Programmable Gate Arrays (FPGA) have been used in many applications to achieve orders-of-magnitude improvement in absolute performance and energy efficiency relative to conventional microprocessors. Despite their newfound potency in both processing performance and energy efficiency, FPGAs have not gained widespread acceptance as mainstream computing devices. A fundamental obstacle to FPGA-based computing can be traced to the FPGA's lack of a common, scalable memory abstraction. When developing for FPGAs, application writers are often responsible for crafting the application-specific infrastructure logic that transports the data to and from the processing kernels, which are the ultimate producers and consumers within the fabric. Very often, this infrastructure logic not only increases design time and effort but will inflexibly lock a design to a particular FPGA product line, hindering scalability and portability. To create a common, scalable memory abstraction, this thesis proposes a new FPGA memory architecture called Connected RAM (CoRAM) to serve as a portable bridge between the distributed computation kernels and the edge memory interfaces. In addition to improving performance and efficiency, the CoRAM architecture provides a virtualized memory environment as seen by the hardware kernels to simplify application development and to improve an application's scalability and portability
展开▼
