An approximation algorithm composed of a digital neural network (DNN) and a modified greedy algorithm (MGA) is presented for the board-level routing problem (BLRP) in a logic emulation system based on field-programmable gate arrays (FPGA's) in this paper. For a rapid prototyping of large scale digital systems, multiple FPGA's provide an efficient logic emulation system, where signals or nets between design partitions embedded on different FPGA's are connected through crossbars. The goal of BLRP, known to be NP-complete in general, is to find a net assignment to crossbars subject to the constraint that all the terminals of any net must be connected through a single crossbar while the number of I/O pins designated for each crossbar m is limited in an FPGA. In the proposed combination algorithm, DNN is applied for m = 1 and MGA is for m 2 in order to achieve the high solution quality. The DNN for the N-net-M-crossbar BLRP consists of N M digital neurons of binary outputs and range-limited non-negative integer inputs with integer parameters. The MGA is modified from the algorithm by Lin et al. The performance is verified through massive simulations, where our algorithm drastically improves the routing capability over the latest greedy algorithms.

The paper describes the architecture and the simulated performances of a memory-based chip that emulates human cortical processing in early visual tasks, such as texture segregation. The featural elements present in an image are extracted by a convolution block and subsequently processed by the cortical chip, whose neurons, organized into three layers, gain relational descriptions (intelligent processing) through recurrent inhibitory/excitatory interactions between both inter-and intra-layer parallel pathways. The digital implementation of this architecuture directly maps the set of equations determining the status of the cortical network to achieve an optimal exploitation of VLSI technology in neural computation. Neurons are mapped into a memory matrix whose elements are updated through a programmable computational unit that implements synaptic interconnections. By using 0.5 µm-CMOS technology, full cortical image processing can be attained on a single chip (2020 mm2 die) at a rate higher than 70 frames/second, for images of 256256 pixels.