Path lookahead: a data flow view of PDES models

  • Authors:

  • Richard A. Meyer;Rajive L. Bargrodia

  • Affiliations:

  • Computer Science Department, University of California at Los Angeles;Computer Science Department, University of California at Los Angeles

  • Venue:
  • PADS '99 Proceedings of the thirteenth workshop on Parallel and distributed simulation
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

Architectural advances are making PDES more difficult. Processor speed improves much more quickly than interprocessor communication and memory. Lagging memory performance has a much greater impact on optimistic techniques, especially for large scale models, but conservative techniques require more small protocol messages and are therefore impacted by slow IPC. The performance of conservative PDES is heavily dependent on the lookahead available in the simulation model. Finding larger lookahead not only allows for increased parallelism, it also reduces the number of protocol messages required. In this paper, a global view of a PDES model as a set of data flows is presented. Using this view, the lookahead of the model can be optimized, resulting in a significant decrease in protocol messages, with only a marginal increase in computation, using realistic, detailed models as examples.