Processor array self-reconfiguration by neural networks

The authors introduce a novel type of neural network which can be intelligently employed for controlling the reconfiguration circuits within a VLSI/WSI chip. In this implementation, the neural network is interconnected and programmed such that it can readily execute a maximum matching algorithm in order to assign fault-free spare elements to faulty components. This approach has been compared with the traditional reconfiguration algorithms, and by intensive simulation it is shown that the proposed neural net approach provides superior quality performance (i.e., higher survivability rates). It is also shown that the intrinsic fault-tolerant nature of neural networks maintains a degradable reconfiguration control even in the presence of faulty neural network components. The speed of neural networks provides an added advantage for online reconfiguration, where the chip can be quickly repaired by itself, thus reducing the system down-time