Targeting GPUs with OpenMP directives on Summit: A simple and effective Fortran experience

Reuben D. Budiardja; Christian Y. Cardall

doi:10.1016/j.parco.2019.102544

Targeting GPUs with OpenMP directives on Summit: A simple and effective Fortran experience

Reuben D. Budiardja, Christian Y. Cardall

University of Tennessee, Knoxville

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

10 Scopus citations

Abstract

We use OpenMP to target hardware accelerators (GPUs) on Summit, a newly deployed supercomputer at the Oak Ridge Leadership Computing Facility (OLCF), demonstrating simplified access to GPU devices for users of our astrophysics code GENASIS and useful speedup on a sample fluid dynamics problem. We modify our workhorse class for data storage to include members and methods that significantly streamline the persistent allocation of and association to GPU memory. Users offload computational kernels with OpenMP target directives that are rather similar to constructs already familiar from multi-core parallelization. In this initial example we ask, “With a given number of Summit nodes, how fast can we compute with and without GPUs?”, and find total wall time speedups of ∼ 12X. We also find reasonable weak scaling up to 8000 GPUs (1334 Summit nodes). We make available the source code from this work at https://github.com/GenASiS/GenASiS_Basics.

Original language	English
Article number	102544
Journal	Parallel Computing
Volume	88
DOIs	https://doi.org/10.1016/j.parco.2019.102544
State	Published - Oct 2019

Bibliographical note

Publisher Copyright:
© 2019 Elsevier B.V.

Funding

This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Nuclear Physics under contract number DE-AC05-00OR22725; the National Science Foundation under Grant No. 1535130 . This research used resources of the Oak Ridge Leadership Computing Facility, which is a DOE Office of Science User Facility supported under Contract DE-AC05-00OR22725. This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Nuclear Physics under contract number DE-AC05-00OR22725; the National Science Foundation under Grant No. 1535130. This research used resources of the Oak Ridge Leadership Computing Facility, which is a DOE Office of Science User Facility supported under Contract DE-AC05-00OR22725.

Funders	Funder number
DOE Office of Science
DOE Office of Science User Facility supported
Oak
Office of Nuclear Physics	DE-AC05-00OR22725
National Science Foundation	1535130
U.S. Department of Energy
Office of Science

Access to Document

10.1016/j.parco.2019.102544

Cite this

@article{2dfbc734b48242459437f6709a76f5e2,

title = "Targeting GPUs with OpenMP directives on Summit: A simple and effective Fortran experience",

abstract = "We use OpenMP to target hardware accelerators (GPUs) on Summit, a newly deployed supercomputer at the Oak Ridge Leadership Computing Facility (OLCF), demonstrating simplified access to GPU devices for users of our astrophysics code GENASIS and useful speedup on a sample fluid dynamics problem. We modify our workhorse class for data storage to include members and methods that significantly streamline the persistent allocation of and association to GPU memory. Users offload computational kernels with OpenMP target directives that are rather similar to constructs already familiar from multi-core parallelization. In this initial example we ask, “With a given number of Summit nodes, how fast can we compute with and without GPUs?”, and find total wall time speedups of ∼ 12X. We also find reasonable weak scaling up to 8000 GPUs (1334 Summit nodes). We make available the source code from this work at https://github.com/GenASiS/GenASiS_Basics.",

author = "Budiardja, {Reuben D.} and Cardall, {Christian Y.}",

note = "Publisher Copyright: {\textcopyright} 2019 Elsevier B.V.",

year = "2019",

month = oct,

doi = "10.1016/j.parco.2019.102544",

language = "English",

volume = "88",

journal = "Parallel Computing",

issn = "0167-8191",

publisher = "Elsevier",

}

TY - JOUR

T1 - Targeting GPUs with OpenMP directives on Summit

T2 - A simple and effective Fortran experience

AU - Budiardja, Reuben D.

AU - Cardall, Christian Y.

PY - 2019/10

Y1 - 2019/10

N2 - We use OpenMP to target hardware accelerators (GPUs) on Summit, a newly deployed supercomputer at the Oak Ridge Leadership Computing Facility (OLCF), demonstrating simplified access to GPU devices for users of our astrophysics code GENASIS and useful speedup on a sample fluid dynamics problem. We modify our workhorse class for data storage to include members and methods that significantly streamline the persistent allocation of and association to GPU memory. Users offload computational kernels with OpenMP target directives that are rather similar to constructs already familiar from multi-core parallelization. In this initial example we ask, “With a given number of Summit nodes, how fast can we compute with and without GPUs?”, and find total wall time speedups of ∼ 12X. We also find reasonable weak scaling up to 8000 GPUs (1334 Summit nodes). We make available the source code from this work at https://github.com/GenASiS/GenASiS_Basics.

AB - We use OpenMP to target hardware accelerators (GPUs) on Summit, a newly deployed supercomputer at the Oak Ridge Leadership Computing Facility (OLCF), demonstrating simplified access to GPU devices for users of our astrophysics code GENASIS and useful speedup on a sample fluid dynamics problem. We modify our workhorse class for data storage to include members and methods that significantly streamline the persistent allocation of and association to GPU memory. Users offload computational kernels with OpenMP target directives that are rather similar to constructs already familiar from multi-core parallelization. In this initial example we ask, “With a given number of Summit nodes, how fast can we compute with and without GPUs?”, and find total wall time speedups of ∼ 12X. We also find reasonable weak scaling up to 8000 GPUs (1334 Summit nodes). We make available the source code from this work at https://github.com/GenASiS/GenASiS_Basics.

UR - http://www.scopus.com/inward/record.url?scp=85071573048&partnerID=8YFLogxK

U2 - 10.1016/j.parco.2019.102544

DO - 10.1016/j.parco.2019.102544

M3 - Article

AN - SCOPUS:85071573048

SN - 0167-8191

VL - 88

JO - Parallel Computing

JF - Parallel Computing

M1 - 102544

ER -

Targeting GPUs with OpenMP directives on Summit: A simple and effective Fortran experience

Abstract

Bibliographical note

Funding

Access to Document

Other files and links

Fingerprint

Cite this