Fine-grained parallelization of the Car-Parrinello ab initio molecular dynamics method on the IBM Blue Gene/L supercomputer

Eric Bohm, Abhinav Bhatele, Laxmikant V. Kalé, Mark Tuckerman, Sameer Kumar, John A. Gunnels, Glenn J. Martyna

Research output: Contribution to journalArticle

Abstract

Important scientific problems can be treated via ab initio-based molecular modeling approaches, wherein atomic forces are derived from an energy function that explicitly considers the electrons. The Car-Parrinello ab initio molecular dynamics (CPAIMD) method is widely used to study small systems containing on the order of 10 10 103 atoms. However, the impact of CPAIMD has been limited until recently because of difficulties inherent to scaling the technique beyond processor numbers about equal to the number of electronic states. CPAIMD computations involve a large number of interdependent phases with high interprocessor communication overhead. These phases require the evaluation of various transforms and non-square matrix multiplications that require large interprocessor data movement when efficiendy parallelized. Using the Charm-++ parallel programming language and runtime system, the phases are discretized into a large number of virtual processors, which are, in turn, mapped flexibly onto physical processors, thereby allowing interleaving of work. Algorithmic and IBM Blue Gene/L™ system-specific optimizations are employed to scale the CPAIMD method to at least 30 times the number of electronic states in small systems consisting of 24 to 768 atoms (32 to 1,024 electronic states) in order to demonstrate fine-grained parallelism. The largest systems studied scaled well across the entire machine (20,480 nodes).

Original languageEnglish (US)
Pages (from-to)159-176
Number of pages18
JournalIBM Journal of Research and Development
Volume52
Issue number1-2
DOIs
StatePublished - 2008

Fingerprint

Supercomputers
Molecular dynamics
Electronic states
Railroad cars
Genes
Atoms
Parallel programming
Molecular modeling
Computer programming languages
Electrons
Communication

ASJC Scopus subject areas

  • Hardware and Architecture

Cite this

Fine-grained parallelization of the Car-Parrinello ab initio molecular dynamics method on the IBM Blue Gene/L supercomputer. / Bohm, Eric; Bhatele, Abhinav; Kalé, Laxmikant V.; Tuckerman, Mark; Kumar, Sameer; Gunnels, John A.; Martyna, Glenn J.

In: IBM Journal of Research and Development, Vol. 52, No. 1-2, 2008, p. 159-176.

Research output: Contribution to journalArticle

Bohm, Eric ; Bhatele, Abhinav ; Kalé, Laxmikant V. ; Tuckerman, Mark ; Kumar, Sameer ; Gunnels, John A. ; Martyna, Glenn J. / Fine-grained parallelization of the Car-Parrinello ab initio molecular dynamics method on the IBM Blue Gene/L supercomputer. In: IBM Journal of Research and Development. 2008 ; Vol. 52, No. 1-2. pp. 159-176.
@article{8747f6ae98c54d449af1cb2c4066d0f0,
title = "Fine-grained parallelization of the Car-Parrinello ab initio molecular dynamics method on the IBM Blue Gene/L supercomputer",
abstract = "Important scientific problems can be treated via ab initio-based molecular modeling approaches, wherein atomic forces are derived from an energy function that explicitly considers the electrons. The Car-Parrinello ab initio molecular dynamics (CPAIMD) method is widely used to study small systems containing on the order of 10 10 103 atoms. However, the impact of CPAIMD has been limited until recently because of difficulties inherent to scaling the technique beyond processor numbers about equal to the number of electronic states. CPAIMD computations involve a large number of interdependent phases with high interprocessor communication overhead. These phases require the evaluation of various transforms and non-square matrix multiplications that require large interprocessor data movement when efficiendy parallelized. Using the Charm-++ parallel programming language and runtime system, the phases are discretized into a large number of virtual processors, which are, in turn, mapped flexibly onto physical processors, thereby allowing interleaving of work. Algorithmic and IBM Blue Gene/L™ system-specific optimizations are employed to scale the CPAIMD method to at least 30 times the number of electronic states in small systems consisting of 24 to 768 atoms (32 to 1,024 electronic states) in order to demonstrate fine-grained parallelism. The largest systems studied scaled well across the entire machine (20,480 nodes).",
author = "Eric Bohm and Abhinav Bhatele and Kal{\'e}, {Laxmikant V.} and Mark Tuckerman and Sameer Kumar and Gunnels, {John A.} and Martyna, {Glenn J.}",
year = "2008",
doi = "10.1147/rd.521.0159",
language = "English (US)",
volume = "52",
pages = "159--176",
journal = "IBM Journal of Research and Development",
issn = "0018-8646",
publisher = "IBM Corporation",
number = "1-2",

}

TY - JOUR

T1 - Fine-grained parallelization of the Car-Parrinello ab initio molecular dynamics method on the IBM Blue Gene/L supercomputer

AU - Bohm, Eric

AU - Bhatele, Abhinav

AU - Kalé, Laxmikant V.

AU - Tuckerman, Mark

AU - Kumar, Sameer

AU - Gunnels, John A.

AU - Martyna, Glenn J.

PY - 2008

Y1 - 2008

N2 - Important scientific problems can be treated via ab initio-based molecular modeling approaches, wherein atomic forces are derived from an energy function that explicitly considers the electrons. The Car-Parrinello ab initio molecular dynamics (CPAIMD) method is widely used to study small systems containing on the order of 10 10 103 atoms. However, the impact of CPAIMD has been limited until recently because of difficulties inherent to scaling the technique beyond processor numbers about equal to the number of electronic states. CPAIMD computations involve a large number of interdependent phases with high interprocessor communication overhead. These phases require the evaluation of various transforms and non-square matrix multiplications that require large interprocessor data movement when efficiendy parallelized. Using the Charm-++ parallel programming language and runtime system, the phases are discretized into a large number of virtual processors, which are, in turn, mapped flexibly onto physical processors, thereby allowing interleaving of work. Algorithmic and IBM Blue Gene/L™ system-specific optimizations are employed to scale the CPAIMD method to at least 30 times the number of electronic states in small systems consisting of 24 to 768 atoms (32 to 1,024 electronic states) in order to demonstrate fine-grained parallelism. The largest systems studied scaled well across the entire machine (20,480 nodes).

AB - Important scientific problems can be treated via ab initio-based molecular modeling approaches, wherein atomic forces are derived from an energy function that explicitly considers the electrons. The Car-Parrinello ab initio molecular dynamics (CPAIMD) method is widely used to study small systems containing on the order of 10 10 103 atoms. However, the impact of CPAIMD has been limited until recently because of difficulties inherent to scaling the technique beyond processor numbers about equal to the number of electronic states. CPAIMD computations involve a large number of interdependent phases with high interprocessor communication overhead. These phases require the evaluation of various transforms and non-square matrix multiplications that require large interprocessor data movement when efficiendy parallelized. Using the Charm-++ parallel programming language and runtime system, the phases are discretized into a large number of virtual processors, which are, in turn, mapped flexibly onto physical processors, thereby allowing interleaving of work. Algorithmic and IBM Blue Gene/L™ system-specific optimizations are employed to scale the CPAIMD method to at least 30 times the number of electronic states in small systems consisting of 24 to 768 atoms (32 to 1,024 electronic states) in order to demonstrate fine-grained parallelism. The largest systems studied scaled well across the entire machine (20,480 nodes).

UR - http://www.scopus.com/inward/record.url?scp=40749125697&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=40749125697&partnerID=8YFLogxK

U2 - 10.1147/rd.521.0159

DO - 10.1147/rd.521.0159

M3 - Article

AN - SCOPUS:40749125697

VL - 52

SP - 159

EP - 176

JO - IBM Journal of Research and Development

JF - IBM Journal of Research and Development

SN - 0018-8646

IS - 1-2

ER -