The advantages of attaining a high computational throughput for numerically intensive digital algorithms through highly parallel distributed processors are described. This approach is shown to be especially appropriate for signal processing rather than more general forms of computing. A machine architecture which exploits this is in course of development and algorithms for real-time processing in radar, speech recognition, image processing and ESM are being developed and benchmarked. The results indicate that the capability of a fully programmable nonspecialised processor of this type compares well with that of contemporary dedicated hardware and promises to exploit small geometry VLSI better than less regular designs of hardware.