Keywords: Dataflow, Synchronous dataflow, multidimensional dataflow, multidimensional signal processing with non-rectangular lattices, graphical programming environments, block diagram languages, tools for rapid prototyping.

A Generalization of Multidimensional Synchronous Dataflow to Handle Arbitrary Sampling Lattices

UCB Tech. Memorandum UCB/ERL M95/59, March 20, 1995, Electronics Research Laboratory, UC Berkeley CA 94720.

by Praveen. K. Murthy, and Edward. A. Lee.

ABSTRACT

Multidimensional Synchronous Dataflow (MDSDF) [15][7] is a model of computation that has been pro posed for specifying multidimensional multirate signal processing systems such as image and video processing algo rithms. The model is an extension of synchronous dataflow (SDF) [14] and has all of the desirable properties of the SDF model such as static schedulability, exposition of data and functional parallelism, and a visually pleasing syntax well suited for block diagram signal processing environments such as Ptolemy [6] and Khoros [13]. However, the MDSDF model as specified in [15] is limited to modelling multidimensional systems sampled on the standard rectan gular lattice. Since many multidimensional signals of practical interest are sampled on non-rectangular lattices, for example, 2:1 interlaced video signals, and many multidimensional multirate systems use non-rectangular multirate operators like hexagonal decimators, it is of interest to have models that are capable of representing and simulating such systems. This report describes an extension of the MDSDF model that allows signals on arbitrary sampling lat tices to be represented, and that allows the use of non-rectangular downsamplers and upsamplers. [33 pages]

The following is a conference (Proceedings of the ICASSP 1996, Atlanta, GA, May 7-10 1996) version of the above. It is shorter and contains only the main ideas.