While the architecture of present-day parallel supercomputers is largely based on the concept of a shared memory, with its attendant limitations of common access, advances in semicoductor technology have led to the development of highly parellel computer architectures with decentralized storage and limited connections in which each processor possesses high bandwidth local memory connected to a small number of such architectures, enabling cost-effective high-speed parallel processing for large volumes of data, with ultra-high throughput rates. Algorithms suitable for implementation on systolic arrays find applications in areas such as signal and image processing, pattern matching, linear algebra, recurrence algorithms and graph problems. This book provides an insight into the implementation of systolic arrays and gives a comprehensive overview of the techniques and theories contributing to the design of systolic algorithms.
