Joint Minimization of Code and Data for Synchronous DataflowPrograms

  • Authors:

  • Praveen K. Murthy;Shuvra S. Bhattacharyya;Edward A. Lee

  • Affiliations:

  • Dept. of Electrical Engineering and Computer Sciences, University of California at Berkeley, California 94720, USA;Semiconductor Research Laboratory, Hitachi America, Ltd., 201 East Tasman Drive, San Jose, California 95134, USA;Dept. of Electrical Engineering and Computer Sciences, University of California at Berkeley, California 94720, USA

  • Venue:
  • Formal Methods in System Design
  • Year:
  • 1997

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, we formally develop techniques that minimize thememory requirements of a target program when synthesizing softwarefrom dataflow descriptions of multirate signal processingalgorithms. The dataflow programming model that we consider is thesynchronous dataflow (SDF)model [21], which has been used heavily in DSP design environmentsover the past several years. We first focus on the restricted classof well-ordered SDF graphs. We show that whileextremely efficient techniques exist for constructing minimum codesize schedules for well-ordered graphs, the number of distinctminimum code size schedules increases combinatorially with the numberof vertices in the input SDF graph, and these different schedules canhave vastly different data memory requirements. We develop a dynamicprogramming algorithm that computes the schedule that minimizes thedata memory requirement from among the schedules that minimize codesize, and we show that the time complexity of this algorithm is cubicin the number of vertices in the given well-ordered SDF graph. Wepresent several extensions to this dynamic programming technique tomore general scheduling problems, and we present a heuristic thatoften computes near-optimal schedules with quadratic time complexity.We then show that finding optimal solutions for arbitrary acyclicgraphs is NP-complete, and present heuristic techniques that jointlyminimize code and data size requirements. We present a practicalexample and simulation data that demonstrate the effectiveness ofthese techniques.