A Fine-grain Dataflow Model for Scientific Visualization Systems

Dataaow computing is a simple yet powerful mechanism for constructing distributed visualization applications. Most of the current dataaow based visualization systems use coarse-grain computing models. A major problem of the coarse-grain visualization systems is that they introduce multiple copies of large data sets during execution. In this paper we propose a ne-grain method called the Syntax-directed Dataaow Transformation Method that combines the dataaow programming paradigm with the syntax-directed translation techniques. Our new model extends the classical dataaow model in which tokens and nodal functions are atomic. It allows structural deenition of the tokens and explicitly states the sequence of nodal actions in relation to the token structures. A dataaow system based on the SDTM model achieves better memory eeciency by keeping a dynamically adjusted minimal window on the input data stream, through the use of syntax-directed translation (attribute grammar) to specify data transformation. CPU utilization can also be improved on a multiprocessor system. A prototype implementation showed a speedup of more than two times over a typical coarse-grain dataaow system and a dramatic reduction in memory usage.