Prerouted FPGA Cores for Rapid System Construction in a Dynamic Reconfigurable System

摘要

A method of constructing prerouted FPGA cores, which lays the foundations for a rapid system construction framework for dynamically reconfigurable computing systems, is presented. Two major challenges are considered: how to manage the wires crossing a core's borders; and how to maintain an acceptable level of flexibility for system construction with only a minimum of overhead. In order to maintain FPGA computing performance, it is crucial to thoroughly analyze the issues at the lowest level of device detail in order to ensure that computing circuit encapsulation is as efficient as possible. We present the first methodology that allows a core to scale its interface bandwidth to the maximum available in a routing channel. Cores can be constructed independently from the rest of the system using a framework that is independent of the method used to place and route primitive components within the core. We use an abstract FPGA model and CAD tools that mirror those used in industry. An academic design flow has been modified to include a wire policy and an interface constraints framework that tightly constrains the use of the wires that cross a core's boundaries. Using this tool set we investigate the effect of prerouting on overall system optimality. Abutting cores are instantly connected by colocation of interface wires. Eliminating run-time routing drastically reduces the time taken to construct a system using a set of cores.