Running multiple applications optimally in terms of Quality of Service (e.g. performance and power consumption) on embedded multi-core platforms is a huge challenge. Moreover, current applications exhibit unpredictable changes of the environment and workload conditions which makes the task of running them optimally even more difficult. This dynamic trend in application runs will grow even more in future applications. This Chapter presents an automated tool flow which tackles this challenge by a two-step approach: first at design-time, a Design Space Exploration (DSE) tool is coupled with a platform simulator(s) to get optimum operating points for the set of target applications. Secondly, at run-time, a lightweight Run-time Resource Manager (RRM) leverages the design-timeDSEresults for deciding an operating configuration to be loaded at run-time for each application. This decision is performed dynamically, by taking into consideration available platform resources and the QoS requirements of the specific use-case. To keepRRMexecution and resource overhead at minimum, a very fast optimisation heuristic is integrated. Application of this tool-flowon a real-life multimedia use case will demonstrate a significant speedup in optimisation process while maintaining desired Quality of Service.
Design Space Exploration Supporting Run-time Resource Management
PALERMO, GIANLUCA;SILVANO, CRISTINA;ZACCARIA, VITTORIO
2011-01-01
Abstract
Running multiple applications optimally in terms of Quality of Service (e.g. performance and power consumption) on embedded multi-core platforms is a huge challenge. Moreover, current applications exhibit unpredictable changes of the environment and workload conditions which makes the task of running them optimally even more difficult. This dynamic trend in application runs will grow even more in future applications. This Chapter presents an automated tool flow which tackles this challenge by a two-step approach: first at design-time, a Design Space Exploration (DSE) tool is coupled with a platform simulator(s) to get optimum operating points for the set of target applications. Secondly, at run-time, a lightweight Run-time Resource Manager (RRM) leverages the design-timeDSEresults for deciding an operating configuration to be loaded at run-time for each application. This decision is performed dynamically, by taking into consideration available platform resources and the QoS requirements of the specific use-case. To keepRRMexecution and resource overhead at minimum, a very fast optimisation heuristic is integrated. Application of this tool-flowon a real-life multimedia use case will demonstrate a significant speedup in optimisation process while maintaining desired Quality of Service.File | Dimensione | Formato | |
---|---|---|---|
CH5_Pre_print_978-1-4419-8837-9_Book_ProofPDF_M3.pdf
Accesso riservato
:
Pre-Print (o Pre-Refereeing)
Dimensione
219.84 kB
Formato
Adobe PDF
|
219.84 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.