Typical high-end DSP systems require a total workload of the order of a few thousand MIPS.Comparing this figure with the 50-100 MIPS average performance of modern DSP chips,one often has to use even up to 100 processors in a single system.Thus methods of implementing parallel processing algorithms become very important.This is the case especially in such systems,where processing nodes have assigned different tasks and thus there exists complicated and extensive data flow between them.
展开▼
