Evaluation des performances pour les systèmes embarqués hétérogènes, multiprocesseur monopuces

Abstract : Multi-processor system-on-chip (MPSoC) is a concept that aims at integrating multiple subsystems on a single chip. Systems that put together complex HW and SW subsystems are difficult to analyze and even harder to optimize. Performance evaluation is a key step in any design, allowing for decisions and trade-offs, in view of overall system optimization. The literature shows that a large part of the design time spent in performance evaluation, and iterations become prohibitive in complex designs. Therefore, the challenge of building high-performance MPSoCs is closely related to the availability of fast and accurate performance evaluation methods. In our work, “performance” is restricted to time related performances of the final architecture. The timing aspect is intensively analyzed for the validation of real-time systems and the optimization of interconnect subsystems. We are also concerned with the speed of any proposed performance evaluation method, as evaluation times may become prohibitive for complex MPSoC designs. Our main objective is to define a global performance evaluation methodology for MPSoC. We also orient our research towards software performance modeling, maintaining a high level of abstraction, in order to have a high evaluation speed, and including timing annotations, in order to have good evaluation
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00086762
Contributor : Lucie Torella <>
Submitted on : Wednesday, July 19, 2006 - 3:32:48 PM
Last modification on : Wednesday, May 16, 2018 - 6:30:04 PM
Long-term archiving on : Tuesday, April 6, 2010 - 12:18:19 AM

Identifiers

  • HAL Id : tel-00086762, version 1

Collections

TIMA | UGA

Citation

I. Bacivarov. Evaluation des performances pour les systèmes embarqués hétérogènes, multiprocesseur monopuces. Micro et nanotechnologies/Microélectronique. Institut National Polytechnique de Grenoble - INPG, 2006. Français. ⟨tel-00086762⟩

Share

Metrics

Record views

382

Files downloads

1301