A Framework for Characterizing Predictable Platform Templates
2014 (English)Report (Other academic)
The design of real-time multiprocessor systems is a very costly and time-consuming process due to the need for extensive verification efforts. Genericcorrect-by-construction system-level design flows, targeting predictable plat-forms, would help to tackle this problem. Unfortunately, because system-level design problems are formulated monolithically, existing methods areeither not powerful enough to perform efficient design space exploration,over-customized to a specific class of platforms, or do not allow to be ex-tended with new heuristics and solving methods, which makes their reusedifficult. We present a formal framework to explicitly capture and character-ize predictable platform templates that can be used to formulate a genericdesign flow for real-time streaming applications in a composable manner. Aproof-of-concept implementation of such a flow is performed and used to mapa JPEG encoder application onto an FPGA-based time-predictable platform.
Place, publisher, year, edition, pages
Stockholm, Sweden: KTH Royal Institute of Technology, 2014. , 18 p.
, TRITA-ICT/ECS R, ISSN 1653-7238 ; 14:01
automation, design-space exploration, predictable platforms, real-time systems
IdentifiersURN: urn:nbn:se:kth:diva-148162ISRN: KTH/ICT/ECS/R-14-01-SEOAI: oai:DiVA.org:kth-148162DiVA: diva2:735860
QC 201408192014-08-012014-08-012014-11-17Bibliographically approved