CAPLab

PaSTA

SDF graph mapper considering task, data and time parallelizm simultaneously

About

  • PaSTA determines node to processor mapping of SDF graph considering data, task and temporal parallelisms. In contrast to the previous researches, multi-rate task graph with cycles are allowed as input task graph in PaSTA.
  • Since the mapping and scheduling problem is NP-hard, PaSTA uses QEA heuristic to optimize the problem. The mapping information and pipeline information are represented in one Q-string and optimize simultaneously. To enumerate all possible pipelines and finds a valid combination of them, Pipeline Ordering Graph(POG) is used.

Download

Usage

Java -jar PaSTA.jar [-pipe] [-dp] [input]

Publication

  • Hoeseok Yang, Soonhoi Ha, "Pipelined Data Parallel Task Mapping/Scheduling Technique for MPSoC", Design Automation and Test in Europe, pp. 69-74, Apr, 2009.

    • Contributor

  • 하순회(Soonhoi Ha)
  • 양회석(Hoeseok Yang)