Evaluation of dataflow programming models for electronic structure theory

Heike Jagode; Anthony Danalis; Reazul Hoque; Mathieu Faverge; Jack Dongarra

doi:10.1002/cpe.4490

Evaluation of dataflow programming models for electronic structure theory

Heike Jagode, Anthony Danalis, Reazul Hoque, Mathieu Faverge, Jack Dongarra

Research output: Contribution to journal › Article › peer-review

1 Scopus citations

Abstract

Dataflow programming models have been growing in popularity as a means to deliver a good balance between performance and portability in the post-petascale era. In this paper, we evaluate different dataflow programming models for electronic structure methods and compare them in terms of programmability, resource utilization, and scalability. In particular, we evaluate two programming paradigms for expressing scientific applications in a dataflow form: (1) explicit dataflow, where the dataflow is specified explicitly by the developer, and (2) implicit dataflow, where a task scheduling runtime derives the dataflow using per-task data-access information embedded in a serial program. We discuss our findings and present a thorough experimental analysis using methods from the NWChem quantum chemistry application as our case study, and OpenMP, StarPU, and PaRSEC as the task-based runtimes that enable the different forms of dataflow execution. Furthermore, we derive an abstract model to explore the limits of the different dataflow programming paradigms.

Original language	English
Article number	e4490
Journal	Concurrency and Computation: Practice and Experience
Volume	30
Issue number	17
DOIs	https://doi.org/10.1002/cpe.4490
State	Published - Sep 10 2018
Externally published	Yes

Funding

We thank the anonymous reviewers for their improvement suggestions. This material is based upon work supported by the US Air Force Office of Scientific Research under AFOSR Award No. FA9550-12-1-0476 “A Co-Design Approach for Advances in Software and Hardware (CoDAASH)”. A portion of this research was performed using the Environmental Molecular Sciences Laboratory, a DOE Office of Science User Facility sponsored by the Office of Biological and Environmental Research and located at Pacific Northwest National Laboratory.

Funders	Funder number
Air Force Office of Scientific Research	FA9550-12-1-0476
Office of Science
Biological and Environmental Research
Pacific Northwest National Laboratory

Keywords

CCSD
NWChem
OpenMP
PaRSEC
StarPU
coupled cluster methods
dataflow
task-based runtime

Access to Document

10.1002/cpe.4490

Cite this

@article{6303c9cd3eb9437d934c32ec22fbd00b,

title = "Evaluation of dataflow programming models for electronic structure theory",

abstract = "Dataflow programming models have been growing in popularity as a means to deliver a good balance between performance and portability in the post-petascale era. In this paper, we evaluate different dataflow programming models for electronic structure methods and compare them in terms of programmability, resource utilization, and scalability. In particular, we evaluate two programming paradigms for expressing scientific applications in a dataflow form: (1) explicit dataflow, where the dataflow is specified explicitly by the developer, and (2) implicit dataflow, where a task scheduling runtime derives the dataflow using per-task data-access information embedded in a serial program. We discuss our findings and present a thorough experimental analysis using methods from the NWChem quantum chemistry application as our case study, and OpenMP, StarPU, and PaRSEC as the task-based runtimes that enable the different forms of dataflow execution. Furthermore, we derive an abstract model to explore the limits of the different dataflow programming paradigms.",

keywords = "CCSD, NWChem, OpenMP, PaRSEC, StarPU, coupled cluster methods, dataflow, task-based runtime",

author = "Heike Jagode and Anthony Danalis and Reazul Hoque and Mathieu Faverge and Jack Dongarra",

note = "Publisher Copyright: Copyright {\textcopyright} 2018 John Wiley & Sons, Ltd.",

year = "2018",

month = sep,

day = "10",

doi = "10.1002/cpe.4490",

language = "English",

volume = "30",

journal = "Concurrency and Computation: Practice and Experience",

issn = "1532-0626",

publisher = "wiley",

number = "17",

}

TY - JOUR

T1 - Evaluation of dataflow programming models for electronic structure theory

AU - Jagode, Heike

AU - Danalis, Anthony

AU - Hoque, Reazul

AU - Faverge, Mathieu

AU - Dongarra, Jack

PY - 2018/9/10

Y1 - 2018/9/10

N2 - Dataflow programming models have been growing in popularity as a means to deliver a good balance between performance and portability in the post-petascale era. In this paper, we evaluate different dataflow programming models for electronic structure methods and compare them in terms of programmability, resource utilization, and scalability. In particular, we evaluate two programming paradigms for expressing scientific applications in a dataflow form: (1) explicit dataflow, where the dataflow is specified explicitly by the developer, and (2) implicit dataflow, where a task scheduling runtime derives the dataflow using per-task data-access information embedded in a serial program. We discuss our findings and present a thorough experimental analysis using methods from the NWChem quantum chemistry application as our case study, and OpenMP, StarPU, and PaRSEC as the task-based runtimes that enable the different forms of dataflow execution. Furthermore, we derive an abstract model to explore the limits of the different dataflow programming paradigms.

AB - Dataflow programming models have been growing in popularity as a means to deliver a good balance between performance and portability in the post-petascale era. In this paper, we evaluate different dataflow programming models for electronic structure methods and compare them in terms of programmability, resource utilization, and scalability. In particular, we evaluate two programming paradigms for expressing scientific applications in a dataflow form: (1) explicit dataflow, where the dataflow is specified explicitly by the developer, and (2) implicit dataflow, where a task scheduling runtime derives the dataflow using per-task data-access information embedded in a serial program. We discuss our findings and present a thorough experimental analysis using methods from the NWChem quantum chemistry application as our case study, and OpenMP, StarPU, and PaRSEC as the task-based runtimes that enable the different forms of dataflow execution. Furthermore, we derive an abstract model to explore the limits of the different dataflow programming paradigms.

KW - CCSD

KW - NWChem

KW - OpenMP

KW - PaRSEC

KW - StarPU

KW - coupled cluster methods

KW - dataflow

KW - task-based runtime

UR - http://www.scopus.com/inward/record.url?scp=85045842213&partnerID=8YFLogxK

U2 - 10.1002/cpe.4490

DO - 10.1002/cpe.4490

M3 - Article

AN - SCOPUS:85045842213

SN - 1532-0626

VL - 30

JO - Concurrency and Computation: Practice and Experience

JF - Concurrency and Computation: Practice and Experience

IS - 17

M1 - e4490

ER -

Evaluation of dataflow programming models for electronic structure theory

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this