Experimental Evaluation of Probabilistic Execution-Time Modeling and Analysis Methods for SDF Applications on MPSoCs
Résumé
Abstract. Early validation of software running on multi-processor platforms
is fundamental to guarantee that real-time constraints will be fully
met. In the domain of timing analysis probabilistic simulation techniques
tackle the problem of scalability. However, creation of probabilistic SystemC
models remains a dicult task and is not well supported for multiprocessors
systems. In this paper we present a modeling work ow that
will then be used for an experimental evaluation of probabilistic simulation
techniques. For the modeling process a measurement-based approach
is proposed to favor the creation of trustful models. Using this approach,
worst case end-to-end latency can be estimated with approximation close
to 3% and 11% compared to real implementation. The evaluated probabilistic
simulation techniques demonstrate good potential to deliver fast
yet accurate estimations for multi-processor systems.