A stochastic model to predict the routability of field-programmable gate arrays

One area of particular importance is the design of an FPGA routing architecture, which houses the user-programmable switches and wires that are used to interconnect the FPGAs logic resources. Because the routing switches consume significant chip area and introduce propagation delays, the design of the routing architecture greatly influences both the area utilization and speed performance of an FPGA. FPGA routing architectures have already been studied using experimental techniques. This paper describes a stochastic model that facilitates exploration of a wide range of FPGA routing architectures using a theoretical approach. In the stochastic model an FPGA is represented as an N*N array of logic blocks separated by both horizontal and vertical routing channels, similar to a Xilinx FPGA. A circuit to be routed is represented by additional parameters that specify the total number of connections, and each connection's length and trajectory. The stochastic model gives an analytic expression for the routability of the circuit in the FPGA. Practically speaking, routability can be viewed as the likelihood that a circuit can be successfully routed in a given FPGA. The routability predictions from the model are validated by comparing them with the results of a previously published experimental study on FPGA routability.>