In modern signal processing, there are increasing demands for large-volume and high-speed computations, setting the stage for signal processing special purpose computers. At the same time, the advent of VLSI has generated a visible impact on signal processing by offering almost unlimited computing hardware at extremely low cost. These factors combined have had major effect on the rapid up-grading of current signal processors, from the classical (one-dimensional) transversal filters to highly parallel (two-dimensional) array processors. This paper presents the algorithmic and architectural footing for such an evolution. Special emphasis is placed on alleviating the burden of global interconnection and global synchronization. While the Systolic Array offers an elegant solution, avoiding global interconnection by effectively managing local data movements, its inherent global synchronization requirement may damage its practicality in ultra large scale systems. On the other hand, the asynchronous data-driven nature of the Wavefront Array offers a natural solution to this synchronization problem.
|Original language||English (US)|
|Title of host publication||Unknown Host Publication Title|
|Editors||F. Anceau, Einar J. Aas|
|Number of pages||15|
|State||Published - Dec 1 1983|
All Science Journal Classification (ASJC) codes