The OpenACC data model: Preliminary study on its major challenges and implementations

Michael Wolfe; Seyong Lee; Jungwon Kim; Xiaonan Tian; Rengan Xu; Barbara Chapman; Sunita Chandrasekaran

doi:10.1016/j.parco.2018.07.003

The OpenACC data model: Preliminary study on its major challenges and implementations

Michael Wolfe, Seyong Lee, Jungwon Kim, Xiaonan Tian, Rengan Xu, Barbara Chapman, Sunita Chandrasekaran

Programming Systems

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

7 Scopus citations

Abstract

This paper describes how the OpenACC data model is implemented in current OpenACC compilers, ranging from research compilers (OpenUH and OpenARC) to a commercial compiler (the PGI OpenACC compiler). First, we summarize various memory architectures in today's accelerator systems. We then describe details and issues in implementing the OpenACC data model in three different OpenACC compilers. This includes managing page tables, asynchronous data transfers, asynchronous memory allocate and free, host data construct, aliasing on a data directive, reusing device memory, partially present data, and adjacent data. We also discusses ongoing work to manage large, complex dynamic data structures. We measured the present table lookups, device memory allocation, pinned memory allocation, and managed memory in the three OpenACC compilers using eight OpenACC applications (seven from the SPEC ACCEL benchmark suite and a shock-hydrodynamics mini-application called LULESH).

Original language	English
Pages (from-to)	15-27
Number of pages	13
Journal	Parallel Computing
Volume	78
DOIs	https://doi.org/10.1016/j.parco.2018.07.003
State	Published - Oct 2018

Funding

This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Advanced Scientific Computing Research. This research was supported in part by the Exascale Computing Project (17-SC-20-SC), a collaborative effort of the U.S. Department of Energy Office of Science and the National Nuclear Security Administration.

Funders	Funder number
U.S. Department of Energy Office of Science
U.S. Department of Energy
Office of Science
National Nuclear Security Administration
Advanced Scientific Computing Research	17-SC-20-SC

Keywords

Accelerators
Compiler implementations
Data model
OpenACC

Access to Document

10.1016/j.parco.2018.07.003

Cite this

@article{ff654c7c503f4e5eadc25734870b9187,

title = "The OpenACC data model: Preliminary study on its major challenges and implementations",

abstract = "This paper describes how the OpenACC data model is implemented in current OpenACC compilers, ranging from research compilers (OpenUH and OpenARC) to a commercial compiler (the PGI OpenACC compiler). First, we summarize various memory architectures in today's accelerator systems. We then describe details and issues in implementing the OpenACC data model in three different OpenACC compilers. This includes managing page tables, asynchronous data transfers, asynchronous memory allocate and free, host data construct, aliasing on a data directive, reusing device memory, partially present data, and adjacent data. We also discusses ongoing work to manage large, complex dynamic data structures. We measured the present table lookups, device memory allocation, pinned memory allocation, and managed memory in the three OpenACC compilers using eight OpenACC applications (seven from the SPEC ACCEL benchmark suite and a shock-hydrodynamics mini-application called LULESH).",

keywords = "Accelerators, Compiler implementations, Data model, OpenACC",

author = "Michael Wolfe and Seyong Lee and Jungwon Kim and Xiaonan Tian and Rengan Xu and Barbara Chapman and Sunita Chandrasekaran",

note = "Publisher Copyright: {\textcopyright} 2018 Elsevier B.V.",

year = "2018",

month = oct,

doi = "10.1016/j.parco.2018.07.003",

language = "English",

volume = "78",

pages = "15--27",

journal = "Parallel Computing",

issn = "0167-8191",

publisher = "Elsevier",

}

TY - JOUR

T1 - The OpenACC data model

T2 - Preliminary study on its major challenges and implementations

AU - Wolfe, Michael

AU - Lee, Seyong

AU - Kim, Jungwon

AU - Tian, Xiaonan

AU - Xu, Rengan

AU - Chapman, Barbara

AU - Chandrasekaran, Sunita

PY - 2018/10

Y1 - 2018/10

N2 - This paper describes how the OpenACC data model is implemented in current OpenACC compilers, ranging from research compilers (OpenUH and OpenARC) to a commercial compiler (the PGI OpenACC compiler). First, we summarize various memory architectures in today's accelerator systems. We then describe details and issues in implementing the OpenACC data model in three different OpenACC compilers. This includes managing page tables, asynchronous data transfers, asynchronous memory allocate and free, host data construct, aliasing on a data directive, reusing device memory, partially present data, and adjacent data. We also discusses ongoing work to manage large, complex dynamic data structures. We measured the present table lookups, device memory allocation, pinned memory allocation, and managed memory in the three OpenACC compilers using eight OpenACC applications (seven from the SPEC ACCEL benchmark suite and a shock-hydrodynamics mini-application called LULESH).

AB - This paper describes how the OpenACC data model is implemented in current OpenACC compilers, ranging from research compilers (OpenUH and OpenARC) to a commercial compiler (the PGI OpenACC compiler). First, we summarize various memory architectures in today's accelerator systems. We then describe details and issues in implementing the OpenACC data model in three different OpenACC compilers. This includes managing page tables, asynchronous data transfers, asynchronous memory allocate and free, host data construct, aliasing on a data directive, reusing device memory, partially present data, and adjacent data. We also discusses ongoing work to manage large, complex dynamic data structures. We measured the present table lookups, device memory allocation, pinned memory allocation, and managed memory in the three OpenACC compilers using eight OpenACC applications (seven from the SPEC ACCEL benchmark suite and a shock-hydrodynamics mini-application called LULESH).

KW - Accelerators

KW - Compiler implementations

KW - Data model

KW - OpenACC

UR - http://www.scopus.com/inward/record.url?scp=85050455249&partnerID=8YFLogxK

U2 - 10.1016/j.parco.2018.07.003

DO - 10.1016/j.parco.2018.07.003

M3 - Article

AN - SCOPUS:85050455249

SN - 0167-8191

VL - 78

SP - 15

EP - 27

JO - Parallel Computing

JF - Parallel Computing

ER -

The OpenACC data model: Preliminary study on its major challenges and implementations

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this