Organization of pipeline operations in mapping unit of the dataflow parallel computing system

摘要

The paper reviews the architecture of dataflow parallel computing system, describes the operation mechanisms of one of the main system modules - the mapping unit. The pipeline mode of mapping unit is considered; an increasing of the capacity associated with the introduction of the pipeline is estimated. The process of step-by-step optimization of pipeline with a description of the bottlenecks and ways to bypass them is shown.