Perfect strong scaling using no additional energy

James Demmel; Andrew Gearhart; Benjamin Lipshitz; Oded Schwartz

doi:10.1109/IPDPS.2013.32

Perfect strong scaling using no additional energy

James Demmel, Andrew Gearhart, Benjamin Lipshitz, Oded Schwartz

Research output: Contribution to conference › Paper › peer-review

20 Scopus citations

Abstract

Energy efficiency of computing devices has become a dominant area of research interest in recent years. Most previous work has focused on architectural techniques to improve power and energy efficiency, only a few consider saving energy at the algorithmic level. We prove that a region of perfect strong scaling in energy exists for matrix multiplication (classical and Strassen) and the direct n-body problem via the use of algorithms that use all available memory to replicate data. This means that we can increase the number of processors by some factor and decrease the runtime (both computation and communication) by the same factor, without changing the total energy use.

Original language	English
Pages	649-660
Number of pages	12
DOIs	https://doi.org/10.1109/IPDPS.2013.32
State	Published - 2013
Externally published	Yes
Event	27th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2013 - Boston, MA, United States Duration: 20 May 2013 → 24 May 2013

Conference

Conference	27th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2013
Country/Territory	United States
City	Boston, MA
Period	20/05/13 → 24/05/13

Keywords

Communication-avoiding algorithms
Energy efficient algorithms
Energy lower bounds
Power efficiency

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1109/IPDPS.2013.32

Cite this

@conference{9fda3d60dc384a0490831f780941af42,

title = "Perfect strong scaling using no additional energy",

abstract = "Energy efficiency of computing devices has become a dominant area of research interest in recent years. Most previous work has focused on architectural techniques to improve power and energy efficiency, only a few consider saving energy at the algorithmic level. We prove that a region of perfect strong scaling in energy exists for matrix multiplication (classical and Strassen) and the direct n-body problem via the use of algorithms that use all available memory to replicate data. This means that we can increase the number of processors by some factor and decrease the runtime (both computation and communication) by the same factor, without changing the total energy use.",

keywords = "Communication-avoiding algorithms, Energy efficient algorithms, Energy lower bounds, Power efficiency",

author = "James Demmel and Andrew Gearhart and Benjamin Lipshitz and Oded Schwartz",

year = "2013",

doi = "10.1109/IPDPS.2013.32",

language = "אנגלית",

pages = "649--660",

note = "27th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2013 ; Conference date: 20-05-2013 Through 24-05-2013",

}

TY - CONF

T1 - Perfect strong scaling using no additional energy

AU - Demmel, James

AU - Gearhart, Andrew

AU - Lipshitz, Benjamin

AU - Schwartz, Oded

PY - 2013

Y1 - 2013

N2 - Energy efficiency of computing devices has become a dominant area of research interest in recent years. Most previous work has focused on architectural techniques to improve power and energy efficiency, only a few consider saving energy at the algorithmic level. We prove that a region of perfect strong scaling in energy exists for matrix multiplication (classical and Strassen) and the direct n-body problem via the use of algorithms that use all available memory to replicate data. This means that we can increase the number of processors by some factor and decrease the runtime (both computation and communication) by the same factor, without changing the total energy use.

AB - Energy efficiency of computing devices has become a dominant area of research interest in recent years. Most previous work has focused on architectural techniques to improve power and energy efficiency, only a few consider saving energy at the algorithmic level. We prove that a region of perfect strong scaling in energy exists for matrix multiplication (classical and Strassen) and the direct n-body problem via the use of algorithms that use all available memory to replicate data. This means that we can increase the number of processors by some factor and decrease the runtime (both computation and communication) by the same factor, without changing the total energy use.

KW - Communication-avoiding algorithms

KW - Energy efficient algorithms

KW - Energy lower bounds

KW - Power efficiency

UR - http://www.scopus.com/inward/record.url?scp=84884891973&partnerID=8YFLogxK

U2 - 10.1109/IPDPS.2013.32

DO - 10.1109/IPDPS.2013.32

M3 - ???researchoutput.researchoutputtypes.contributiontoconference.paper???

AN - SCOPUS:84884891973

SP - 649

EP - 660

T2 - 27th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2013

Y2 - 20 May 2013 through 24 May 2013

ER -

Perfect strong scaling using no additional energy

Abstract

Conference

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this