Quantifying the Interaction Between Structural Properties of Software and Hardware in the ARM Big.LITTLE Architecture

A4 Konferenspublikationer


Interna författare/redaktörer


Publikationens författare: Srboljub Stepanovic, Georgios Georgakarakos, Simon Holmbacka, Johan Lilius
Redaktörer: Ivan Merelli, Pietro Lio, Igor Kotenko
Publiceringsår: 2018
Förläggare: IEEE
Moderpublikationens namn: 2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)
Artikelns första sida, sidnummer: 138
Artikelns sista sida, sidnummer: 144
ISBN: 978-1-5386-4976-3
eISBN: 978-1-5386-4975-6
ISSN: 2377-5750


Abstrakt

Heterogeneous architectures offer the opportunity to achieve high performance and energy efficiency by selecting appropriate cores for execution of ever changing software applications. Appropriate core selection depends on the interaction between the structural properties of the software and the hardware that influences performance of the software. We propose a model for efficient core selection when executing software on ARM's big.LITTLE heterogeneous architecture. It features a metric based on the correlation between the performance and the number of last level data cache (LLC) misses on a big and a LITTLE core. Additionally our model defines a soft threshold in terms of the number of LLC misses that determines efficient core selection. We verify the model on both a stress benchmark (stress-ng) and a performance and energy demanding application (HEVC decoding) using XMEM and Linux perf dynamic tools.


Nyckelord

appropriate core selection, ARM big.LITTLE architecture, Benchmark testing, big.LITTLE, Cache storage, cycles per instruction, efficient core selection, energy demanding application, executing software, Hardware, heterogeneous architecture, heterogeneous architectures, HEVC, high performance, last level data cache misses, Linux, Linux perf dynamic tools, LITTLE heterogeneous architecture, LLC, Microarchitecture, microprocessor chips, Multicore processing, multiprocessing systems, parallel architectures, Performance evaluation, Pipelines, power aware computing, Shared memory systems, soft threshold, Software, software applications, stress-ng, Structural properties, Tools, XMEM

Senast uppdaterad 2019-24-10 vid 04:07