Selecting Systolic Designs Using Linear Transformations Of Space-Time
Peter R. Cappello, Kenneth Steiglitz · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1984
Cellular computations (i.e., systolic or wavefront computations) are embedded in a vector space, one of whose dimensions models time, and the others, space. Cellular computations that are related by a linear transformation may have different properties with respect to input/output schedules, chip area, communication topology, latency, period, and the presence/absence of broadcasting and pipelining. Linear transformations of space-time are used to explore design alternatives in a formal way. Cellular computations for convolution and matrix product are used to illustrate this linear transformation technique.