How pre-multicore methods and algorithms perform in multicore era

Alexey Lastovetsky; Muhammad  Fahad; Hamidreza Khaleghzadeh; Semyon Khokhriakov; Ravi Reddy Manumachu; Arsalan  Shahid; Lukasz Szustak; Roman  Wyrzykowski

doi:10.1007/978-3-030-02465-9_37

How pre-multicore methods and algorithms perform in multicore era

Alexey Lastovetsky
, Muhammad Fahad
, Hamidreza Khaleghzadeh
, Semyon Khokhriakov
, Ravi Reddy Manumachu
, Arsalan Shahid
, Lukasz Szustak
, Roman Wyrzykowski

University College Dublin

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Many classical methods and algorithms developed when single-core CPUs dominated the parallel computing landscape, are still widely used in the changed multicore world. Two prominent examples are load balancing, which has been one of the main techniques for minimization of the computation time of parallel applications since the beginning of parallel computing, and model-based power/energy measurement techniques using performance events. In this paper, we show that in the multicore era, load balancing is no longer synonymous to optimization and present recent methods and algorithms for optimization of parallel applications for performance and energy on modern HPC platforms, which do not rely on load balancing and often return imbalanced but optimal solutions.

We also show that some fundamental assumptions about performance events, which have to be true for the model-based power/energy measurement tools to be accurate, are increasingly difficult to satisfy as the number of CPU cores increases. Therefore, energy-aware computing methods relying on these tools will be increasingly difficult to verify.

Original language	English
Title of host publication	High Performance Computing
Subtitle of host publication	ISC High Performance 2018 International Workshops, Frankfurt/Main, Germany, June 28, 2018, Revised Selected Papers
Editors	Rio Yokota, Michèle Weiland, John Shalf, Sadaf Alam
Publisher	Springer
Pages	527-539
ISBN (Electronic)	9783030024659
ISBN (Print)	9783030024642
DOIs	https://doi.org/10.1007/978-3-030-02465-9_37
Publication status	Published - 25 Jan 2019
Event	2018 ISC High Performance Workshops - Frankfurt, Germany Duration: 28 Jun 2018 → …

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
Volume	11203
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	2018 ISC High Performance Workshops
Country/Territory	Germany
City	Frankfurt
Period	28/06/18 → …

Keywords

Multicore platforms
Load balancing
Power and energy modeling
Performance monitoring counters

Access to Document

10.1007/978-3-030-02465-9_37

Cite this

Lastovetsky, A., Fahad, M., Khaleghzadeh, H., Khokhriakov, S., Manumachu, R. R., Shahid, A., Szustak, L., & Wyrzykowski, R. (2019). How pre-multicore methods and algorithms perform in multicore era. In R. Yokota, M. Weiland, J. Shalf, & S. Alam (Eds.), High Performance Computing: ISC High Performance 2018 International Workshops, Frankfurt/Main, Germany, June 28, 2018, Revised Selected Papers (pp. 527-539). (Lecture Notes in Computer Science; Vol. 11203). Springer. https://doi.org/10.1007/978-3-030-02465-9_37

Lastovetsky, Alexey ; Fahad, Muhammad ; Khaleghzadeh, Hamidreza et al. / How pre-multicore methods and algorithms perform in multicore era. High Performance Computing: ISC High Performance 2018 International Workshops, Frankfurt/Main, Germany, June 28, 2018, Revised Selected Papers. editor / Rio Yokota ; Michèle Weiland ; John Shalf ; Sadaf Alam. Springer, 2019. pp. 527-539 (Lecture Notes in Computer Science).

@inproceedings{2148c43854c448fd918aae385280a8c7,

title = "How pre-multicore methods and algorithms perform in multicore era",

abstract = "Many classical methods and algorithms developed when single-core CPUs dominated the parallel computing landscape, are still widely used in the changed multicore world. Two prominent examples are load balancing, which has been one of the main techniques for minimization of the computation time of parallel applications since the beginning of parallel computing, and model-based power/energy measurement techniques using performance events. In this paper, we show that in the multicore era, load balancing is no longer synonymous to optimization and present recent methods and algorithms for optimization of parallel applications for performance and energy on modern HPC platforms, which do not rely on load balancing and often return imbalanced but optimal solutions. We also show that some fundamental assumptions about performance events, which have to be true for the model-based power/energy measurement tools to be accurate, are increasingly difficult to satisfy as the number of CPU cores increases. Therefore, energy-aware computing methods relying on these tools will be increasingly difficult to verify.",

keywords = "Multicore platforms, Load balancing, Power and energy modeling, Performance monitoring counters",

author = "Alexey Lastovetsky and Muhammad Fahad and Hamidreza Khaleghzadeh and Semyon Khokhriakov and Manumachu, \{Ravi Reddy\} and Arsalan Shahid and Lukasz Szustak and Roman Wyrzykowski",

year = "2019",

month = jan,

day = "25",

doi = "10.1007/978-3-030-02465-9\_37",

language = "English",

isbn = "9783030024642",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "527--539",

editor = "Rio Yokota and Mich{\`e}le Weiland and John Shalf and Sadaf Alam",

booktitle = "High Performance Computing",

address = "Netherlands",

note = "2018 ISC High Performance Workshops ; Conference date: 28-06-2018",

}

Lastovetsky, A, Fahad, M, Khaleghzadeh, H, Khokhriakov, S, Manumachu, RR, Shahid, A, Szustak, L & Wyrzykowski, R 2019, How pre-multicore methods and algorithms perform in multicore era. in R Yokota, M Weiland, J Shalf & S Alam (eds), High Performance Computing: ISC High Performance 2018 International Workshops, Frankfurt/Main, Germany, June 28, 2018, Revised Selected Papers. Lecture Notes in Computer Science, vol. 11203, Springer, pp. 527-539, 2018 ISC High Performance Workshops, Frankfurt, Germany, 28/06/18. https://doi.org/10.1007/978-3-030-02465-9_37

How pre-multicore methods and algorithms perform in multicore era. / Lastovetsky, Alexey; Fahad, Muhammad ; Khaleghzadeh, Hamidreza et al.
High Performance Computing: ISC High Performance 2018 International Workshops, Frankfurt/Main, Germany, June 28, 2018, Revised Selected Papers. ed. / Rio Yokota; Michèle Weiland; John Shalf; Sadaf Alam. Springer, 2019. p. 527-539 (Lecture Notes in Computer Science; Vol. 11203).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - How pre-multicore methods and algorithms perform in multicore era

AU - Lastovetsky, Alexey

AU - Fahad, Muhammad

AU - Khaleghzadeh, Hamidreza

AU - Khokhriakov, Semyon

AU - Manumachu, Ravi Reddy

AU - Shahid, Arsalan

AU - Szustak, Lukasz

AU - Wyrzykowski, Roman

PY - 2019/1/25

Y1 - 2019/1/25

N2 - Many classical methods and algorithms developed when single-core CPUs dominated the parallel computing landscape, are still widely used in the changed multicore world. Two prominent examples are load balancing, which has been one of the main techniques for minimization of the computation time of parallel applications since the beginning of parallel computing, and model-based power/energy measurement techniques using performance events. In this paper, we show that in the multicore era, load balancing is no longer synonymous to optimization and present recent methods and algorithms for optimization of parallel applications for performance and energy on modern HPC platforms, which do not rely on load balancing and often return imbalanced but optimal solutions. We also show that some fundamental assumptions about performance events, which have to be true for the model-based power/energy measurement tools to be accurate, are increasingly difficult to satisfy as the number of CPU cores increases. Therefore, energy-aware computing methods relying on these tools will be increasingly difficult to verify.

AB - Many classical methods and algorithms developed when single-core CPUs dominated the parallel computing landscape, are still widely used in the changed multicore world. Two prominent examples are load balancing, which has been one of the main techniques for minimization of the computation time of parallel applications since the beginning of parallel computing, and model-based power/energy measurement techniques using performance events. In this paper, we show that in the multicore era, load balancing is no longer synonymous to optimization and present recent methods and algorithms for optimization of parallel applications for performance and energy on modern HPC platforms, which do not rely on load balancing and often return imbalanced but optimal solutions. We also show that some fundamental assumptions about performance events, which have to be true for the model-based power/energy measurement tools to be accurate, are increasingly difficult to satisfy as the number of CPU cores increases. Therefore, energy-aware computing methods relying on these tools will be increasingly difficult to verify.

KW - Multicore platforms

KW - Load balancing

KW - Power and energy modeling

KW - Performance monitoring counters

U2 - 10.1007/978-3-030-02465-9_37

DO - 10.1007/978-3-030-02465-9_37

M3 - Conference contribution

SN - 9783030024642

T3 - Lecture Notes in Computer Science

SP - 527

EP - 539

BT - High Performance Computing

A2 - Yokota, Rio

A2 - Weiland, Michèle

A2 - Shalf, John

A2 - Alam, Sadaf

PB - Springer

T2 - 2018 ISC High Performance Workshops

Y2 - 28 June 2018

ER -

Lastovetsky A, Fahad M, Khaleghzadeh H, Khokhriakov S, Manumachu RR, Shahid A et al. How pre-multicore methods and algorithms perform in multicore era. In Yokota R, Weiland M, Shalf J, Alam S, editors, High Performance Computing: ISC High Performance 2018 International Workshops, Frankfurt/Main, Germany, June 28, 2018, Revised Selected Papers. Springer. 2019. p. 527-539. (Lecture Notes in Computer Science). doi: 10.1007/978-3-030-02465-9_37

How pre-multicore methods and algorithms perform in multicore era

Abstract

Publication series

Conference

Keywords

Access to Document

Fingerprint

Cite this