Cashmere: Heterogeneous Many-Core Computing

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

Abstract

New generations of many-core hardware become available frequently and are typically attractive extensions for data-centers because of power-consumption and performance benefits. As a result, supercomputers and clusters are becoming heterogeneous and start to contain a variety of many-core devices. Obtaining performance from a homogeneous cluster-computer is already challenging, but achieving it from a heterogeneous cluster is even more demanding. Related work primarily focuses on homogeneous many-core clusters. In this paper we present Cashmere, a programming system for heterogeneous many-core clusters. Cashmere is a tight integration of two existing systems: Satin is a programming system that provides a divide-and-conquer programming model with automatic load-balancing and latency-hiding, while Many-Core Levels is a programming system that provides a powerful methodology to optimize computational kernels for varying types of many-core hardware. We evaluate our system with several classes of applications and show that Cashmere achieves high performance and good scalability. The efficiency of heterogeneous executions is comparable to the homogeneous runs and is >90% in three out of four applications.
LanguageEnglish
Title of host publicationProceedings - 2015 IEEE 29th International Parallel and Distributed Processing Symposium, IPDPS 2015
Place of PublicationHyderabad, India
PublisherInstitute of Electrical and Electronics Engineers, Inc.
Pages135-145
Number of pages11
ISBN (Electronic)9781479986484
DOIs
Publication statusPublished - 17 Jul 2015
Event29th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2015 - Hyderabad, India
Duration: 25 May 201529 May 2015

Conference

Conference29th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2015
CountryIndia
CityHyderabad
Period25/05/1529/05/15

Fingerprint

Computer systems programming
Hardware
Core levels
Supercomputers
Computer programming
Resource allocation
Scalability
Electric power utilization

Keywords

  • cluster
  • divide-and-conquer
  • heterogeneous
  • many-core

Cite this

Hijma, P., Jacobs, C. J. H., van Nieuwpoort, R. V., & Bal, H. E. (2015). Cashmere: Heterogeneous Many-Core Computing. In Proceedings - 2015 IEEE 29th International Parallel and Distributed Processing Symposium, IPDPS 2015 (pp. 135-145). [7161503] Hyderabad, India: Institute of Electrical and Electronics Engineers, Inc.. https://doi.org/10.1109/IPDPS.2015.38
Hijma, Pieter ; Jacobs, Ceriel J.H. ; van Nieuwpoort, Rob V. ; Bal, Henri E. / Cashmere : Heterogeneous Many-Core Computing. Proceedings - 2015 IEEE 29th International Parallel and Distributed Processing Symposium, IPDPS 2015. Hyderabad, India : Institute of Electrical and Electronics Engineers, Inc., 2015. pp. 135-145
@inproceedings{a42584bd5c06453fa14c417d8edabb5b,
title = "Cashmere: Heterogeneous Many-Core Computing",
abstract = "New generations of many-core hardware become available frequently and are typically attractive extensions for data-centers because of power-consumption and performance benefits. As a result, supercomputers and clusters are becoming heterogeneous and start to contain a variety of many-core devices. Obtaining performance from a homogeneous cluster-computer is already challenging, but achieving it from a heterogeneous cluster is even more demanding. Related work primarily focuses on homogeneous many-core clusters. In this paper we present Cashmere, a programming system for heterogeneous many-core clusters. Cashmere is a tight integration of two existing systems: Satin is a programming system that provides a divide-and-conquer programming model with automatic load-balancing and latency-hiding, while Many-Core Levels is a programming system that provides a powerful methodology to optimize computational kernels for varying types of many-core hardware. We evaluate our system with several classes of applications and show that Cashmere achieves high performance and good scalability. The efficiency of heterogeneous executions is comparable to the homogeneous runs and is >90{\%} in three out of four applications.",
keywords = "cluster, divide-and-conquer, heterogeneous, many-core",
author = "Pieter Hijma and Jacobs, {Ceriel J.H.} and {van Nieuwpoort}, {Rob V.} and Bal, {Henri E.}",
year = "2015",
month = "7",
day = "17",
doi = "10.1109/IPDPS.2015.38",
language = "English",
pages = "135--145",
booktitle = "Proceedings - 2015 IEEE 29th International Parallel and Distributed Processing Symposium, IPDPS 2015",
publisher = "Institute of Electrical and Electronics Engineers, Inc.",

}

Hijma, P, Jacobs, CJH, van Nieuwpoort, RV & Bal, HE 2015, Cashmere: Heterogeneous Many-Core Computing. in Proceedings - 2015 IEEE 29th International Parallel and Distributed Processing Symposium, IPDPS 2015., 7161503, Institute of Electrical and Electronics Engineers, Inc., Hyderabad, India, pp. 135-145, 29th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2015, Hyderabad, India, 25/05/15. https://doi.org/10.1109/IPDPS.2015.38

Cashmere : Heterogeneous Many-Core Computing. / Hijma, Pieter; Jacobs, Ceriel J.H.; van Nieuwpoort, Rob V.; Bal, Henri E.

Proceedings - 2015 IEEE 29th International Parallel and Distributed Processing Symposium, IPDPS 2015. Hyderabad, India : Institute of Electrical and Electronics Engineers, Inc., 2015. p. 135-145 7161503.

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Cashmere

T2 - Heterogeneous Many-Core Computing

AU - Hijma, Pieter

AU - Jacobs, Ceriel J.H.

AU - van Nieuwpoort, Rob V.

AU - Bal, Henri E.

PY - 2015/7/17

Y1 - 2015/7/17

N2 - New generations of many-core hardware become available frequently and are typically attractive extensions for data-centers because of power-consumption and performance benefits. As a result, supercomputers and clusters are becoming heterogeneous and start to contain a variety of many-core devices. Obtaining performance from a homogeneous cluster-computer is already challenging, but achieving it from a heterogeneous cluster is even more demanding. Related work primarily focuses on homogeneous many-core clusters. In this paper we present Cashmere, a programming system for heterogeneous many-core clusters. Cashmere is a tight integration of two existing systems: Satin is a programming system that provides a divide-and-conquer programming model with automatic load-balancing and latency-hiding, while Many-Core Levels is a programming system that provides a powerful methodology to optimize computational kernels for varying types of many-core hardware. We evaluate our system with several classes of applications and show that Cashmere achieves high performance and good scalability. The efficiency of heterogeneous executions is comparable to the homogeneous runs and is >90% in three out of four applications.

AB - New generations of many-core hardware become available frequently and are typically attractive extensions for data-centers because of power-consumption and performance benefits. As a result, supercomputers and clusters are becoming heterogeneous and start to contain a variety of many-core devices. Obtaining performance from a homogeneous cluster-computer is already challenging, but achieving it from a heterogeneous cluster is even more demanding. Related work primarily focuses on homogeneous many-core clusters. In this paper we present Cashmere, a programming system for heterogeneous many-core clusters. Cashmere is a tight integration of two existing systems: Satin is a programming system that provides a divide-and-conquer programming model with automatic load-balancing and latency-hiding, while Many-Core Levels is a programming system that provides a powerful methodology to optimize computational kernels for varying types of many-core hardware. We evaluate our system with several classes of applications and show that Cashmere achieves high performance and good scalability. The efficiency of heterogeneous executions is comparable to the homogeneous runs and is >90% in three out of four applications.

KW - cluster

KW - divide-and-conquer

KW - heterogeneous

KW - many-core

UR - http://www.scopus.com/inward/record.url?scp=84971467990&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84971467990&partnerID=8YFLogxK

U2 - 10.1109/IPDPS.2015.38

DO - 10.1109/IPDPS.2015.38

M3 - Conference contribution

SP - 135

EP - 145

BT - Proceedings - 2015 IEEE 29th International Parallel and Distributed Processing Symposium, IPDPS 2015

PB - Institute of Electrical and Electronics Engineers, Inc.

CY - Hyderabad, India

ER -

Hijma P, Jacobs CJH, van Nieuwpoort RV, Bal HE. Cashmere: Heterogeneous Many-Core Computing. In Proceedings - 2015 IEEE 29th International Parallel and Distributed Processing Symposium, IPDPS 2015. Hyderabad, India: Institute of Electrical and Electronics Engineers, Inc. 2015. p. 135-145. 7161503 https://doi.org/10.1109/IPDPS.2015.38