TUTCRIS - Tampereen teknillinen yliopisto

TUTCRIS

Partial expansion graphs: Exposing parallelism and dynamic scheduling opportunities for DSP applications

Tutkimustuotosvertaisarvioitu

Yksityiskohdat

AlkuperäiskieliEnglanti
OtsikkoProceedings - 2012 IEEE 23rd International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2012
Sivut86-93
Sivumäärä8
DOI - pysyväislinkit
TilaJulkaistu - 2012
OKM-julkaisutyyppiA4 Artikkeli konferenssijulkaisussa
Tapahtuma2012 IEEE 23rd International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2012 - Delft, Alankomaat
Kesto: 9 heinäkuuta 201211 heinäkuuta 2012

Conference

Conference2012 IEEE 23rd International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2012
MaaAlankomaat
KaupunkiDelft
Ajanjakso9/07/1211/07/12

Tiivistelmä

Emerging Digital Signal Processing (DSP) algorithms and wireless communications protocols require dynamicadaptation and online reconfiguration for the implementedsystems at runtime. In this paper, we introduce the conceptof Partial Expansion Graphs (PEGs) as an implementationmodel and associated class of scheduling strategies. PEGsare designed to help realize DSP systems in terms of formsand granularities of parallelism that are well matched to thegiven applications and targeted platforms. PEGs also facilitatederivation of both static and dynamic scheduling techniques,depending on the amount of variability in task execution timesand other operating conditions. We show how to implementefficient PEG-based scheduling methods using real time operating systems, and to re-use pre-optimized libraries of DSPcomponents within such implementations. Empirical resultsshow that the PEG strategy can 1) achieve significant speedupson a state of the art multicore signal processor platform forstatic dataflow applications with predictable execution times,and 2) exceed classical scheduling speedups for applicationshaving execution times that can vary dynamically. This abilityto handle variable execution times is especially useful as DSPapplications and platforms increase in complexity and adaptive behavior, thereby reducing execution time predictability.