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

Srboljub Stepanovic; Georgios Georgakarakos; Simon Holmbacka; Johan Lilius

doi:10.1109/PDP2018.2018.00027

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

Srboljub Stepanovic, Georgios Georgakarakos, Simon Holmbacka, Johan Lilius

Forskningsoutput: Kapitel i bok/konferenshandling › Konferensbidrag › Vetenskaplig › Peer review

2 Citeringar (Scopus)

127 Nedladdningar (Pure)

Sammanfattning

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.

Originalspråk	Odefinierat/okänt
Titel på värdpublikation	2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)
Redaktörer	Ivan Merelli, Pietro Lio, Igor Kotenko
Förlag	IEEE
Sidor	138–144
ISBN (elektroniskt)	978-1-5386-4975-6
ISBN (tryckt)	978-1-5386-4976-3
DOI	https://doi.org/10.1109/PDP2018.2018.00027
Status	Publicerad - 2018
MoE-publikationstyp	A4 Artikel i en konferenspublikation
Evenemang	Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP) - Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP) Varaktighet: 21 mars 2018 → 23 mars 2018

Konferens

Konferens	Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)
Period	21/03/18 → 23/03/18

Nyckelord

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

Åtkomst av dokument

10.1109/PDP2018.2018.00027

QuantifyingInteraction.pdfAccepterat manuskript från författare, 403 KBLicens: Publisher rights policy

http://urn.fi/URN:NBN:fi-fe2020102788605

ESC: Efficient Stream Computing by Fitting Computations to Cores
Lilius, J., Lafond, S., Zelioli, L., Iancu, B., Soloviev, V., Georgakarakos, G. & Kanur Chandra Shekar, S.
01/09/17 → 31/08/21
Projekt: Finlands Akademi/Övriga Forskningsråd

Citera det här

Stepanovic, S., Georgakarakos, G., Holmbacka, S., & Lilius, J. (2018). Quantifying the Interaction Between Structural Properties of Software and Hardware in the ARM Big.LITTLE Architecture. I I. Merelli, P. Lio, & I. Kotenko (Red.), 2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP) (s. 138–144). IEEE. https://doi.org/10.1109/PDP2018.2018.00027

Stepanovic, Srboljub ; Georgakarakos, Georgios ; Holmbacka, Simon et al. / Quantifying the Interaction Between Structural Properties of Software and Hardware in the ARM Big.LITTLE Architecture. 2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP). redaktör / Ivan Merelli ; Pietro Lio ; Igor Kotenko. IEEE, 2018. s. 138–144

@inproceedings{9404f048403a4978b9294ad3368dcba1,

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

abstract = "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.",

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

author = "Srboljub Stepanovic and Georgios Georgakarakos and Simon Holmbacka and Johan Lilius",

year = "2018",

doi = "10.1109/PDP2018.2018.00027",

language = "Odefinierat/ok{\"a}nt",

isbn = "978-1-5386-4976-3",

pages = "138–144",

editor = "Ivan Merelli and Pietro Lio and Igor Kotenko",

booktitle = "2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)",

publisher = "IEEE",

note = "Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP) ; Conference date: 21-03-2018 Through 23-03-2018",

}

Stepanovic, S, Georgakarakos, G, Holmbacka, S & Lilius, J 2018, Quantifying the Interaction Between Structural Properties of Software and Hardware in the ARM Big.LITTLE Architecture. i I Merelli, P Lio & I Kotenko (red), 2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP). IEEE, s. 138–144, Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP), 21/03/18. https://doi.org/10.1109/PDP2018.2018.00027

Quantifying the Interaction Between Structural Properties of Software and Hardware in the ARM Big.LITTLE Architecture. / Stepanovic, Srboljub; Georgakarakos, Georgios; Holmbacka, Simon et al.
2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP). red. / Ivan Merelli; Pietro Lio; Igor Kotenko. IEEE, 2018. s. 138–144.

Forskningsoutput: Kapitel i bok/konferenshandling › Konferensbidrag › Vetenskaplig › Peer review

TY - GEN

T1 - Quantifying the Interaction Between Structural Properties of Software and Hardware in the ARM Big.LITTLE Architecture

AU - Stepanovic, Srboljub

AU - Georgakarakos, Georgios

AU - Holmbacka, Simon

AU - Lilius, Johan

PY - 2018

Y1 - 2018

N2 - 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.

AB - 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.

KW - ARM big.LITTLE architecture

KW - Benchmark testing

KW - Cache storage

KW - HEVC

KW - Hardware

KW - LITTLE heterogeneous architecture

KW - LLC

KW - Linux

KW - Linux perf dynamic tools

KW - Microarchitecture

KW - Multicore processing

KW - Performance evaluation

KW - Pipelines

KW - Shared memory systems

KW - Software

KW - Structural properties

KW - Tools

KW - XMEM

KW - appropriate core selection

KW - big.LITTLE

KW - cycles per instruction

KW - efficient core selection

KW - energy demanding application

KW - executing software

KW - heterogeneous architecture

KW - heterogeneous architectures

KW - high performance

KW - last level data cache misses

KW - microprocessor chips

KW - multiprocessing systems

KW - parallel architectures

KW - power aware computing

KW - soft threshold

KW - software applications

KW - stress-ng

KW - ARM big.LITTLE architecture

KW - Benchmark testing

KW - Cache storage

KW - HEVC

KW - Hardware

KW - LITTLE heterogeneous architecture

KW - LLC

KW - Linux

KW - Linux perf dynamic tools

KW - Microarchitecture

KW - Multicore processing

KW - Performance evaluation

KW - Pipelines

KW - Shared memory systems

KW - Software

KW - Structural properties

KW - Tools

KW - XMEM

KW - appropriate core selection

KW - big.LITTLE

KW - cycles per instruction

KW - efficient core selection

KW - energy demanding application

KW - executing software

KW - heterogeneous architecture

KW - heterogeneous architectures

KW - high performance

KW - last level data cache misses

KW - microprocessor chips

KW - multiprocessing systems

KW - parallel architectures

KW - power aware computing

KW - soft threshold

KW - software applications

KW - stress-ng

KW - ARM big.LITTLE architecture

KW - Benchmark testing

KW - Cache storage

KW - HEVC

KW - Hardware

KW - LITTLE heterogeneous architecture

KW - LLC

KW - Linux

KW - Linux perf dynamic tools

KW - Microarchitecture

KW - Multicore processing

KW - Performance evaluation

KW - Pipelines

KW - Shared memory systems

KW - Software

KW - Structural properties

KW - Tools

KW - XMEM

KW - appropriate core selection

KW - big.LITTLE

KW - cycles per instruction

KW - efficient core selection

KW - energy demanding application

KW - executing software

KW - heterogeneous architecture

KW - heterogeneous architectures

KW - high performance

KW - last level data cache misses

KW - microprocessor chips

KW - multiprocessing systems

KW - parallel architectures

KW - power aware computing

KW - soft threshold

KW - software applications

KW - stress-ng

U2 - 10.1109/PDP2018.2018.00027

DO - 10.1109/PDP2018.2018.00027

M3 - Konferensbidrag

SN - 978-1-5386-4976-3

SP - 138

EP - 144

BT - 2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)

A2 - Merelli, Ivan

A2 - Lio, Pietro

A2 - Kotenko, Igor

PB - IEEE

T2 - Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)

Y2 - 21 March 2018 through 23 March 2018

ER -

Stepanovic S, Georgakarakos G, Holmbacka S, Lilius J. Quantifying the Interaction Between Structural Properties of Software and Hardware in the ARM Big.LITTLE Architecture. I Merelli I, Lio P, Kotenko I, redaktörer, 2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP). IEEE. 2018. s. 138–144 doi: 10.1109/PDP2018.2018.00027

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

Sammanfattning

Konferens

Nyckelord

Åtkomst av dokument

Projekt

ESC: Efficient Stream Computing by Fitting Computations to Cores

Citera det här