Optimized Data Transfers Based on the OpenCL Event Management Mechanism

Hiroyuki Takizawa, Shoichi Hirasawa, Makoto Sugawara, Isaac Gelado, Hiroaki Kobayashi, Wen-Mei W Hwu

Research output: Contribution to journalArticle

Abstract

In standard OpenCL programming, hosts are supposed to control their compute devices. Since compute devices are dedicated to kernel computation, only hosts can execute several kinds of data transfers such as internode communication and file access. These data transfers require one host to simultaneously play two or more roles due to the need for collaboration between the host and devices. The codes for such data transfers are likely to be system-specific, resulting in low portability. This paper proposes an OpenCL extension that incorporates such data transfers into the OpenCL event management mechanism. Unlike the current OpenCL standard, the main thread running on the host is not blocked to serialize dependent operations. Hence, an application can easily use the opportunities to overlap parallel activities of hosts and compute devices. In addition, the implementation details of data transfers are hidden behind the extension, and application programmers can use the optimized data transfers without any tricky programming techniques. The evaluation results show that the proposed extension can use the optimized data transfer implementation and thereby increase the sustained data transfer performance by about 18% for a real application accessing a big data file.

Original languageEnglish (US)
Article number576498
JournalScientific Programming
Volume2015
DOIs
StatePublished - Jan 1 2015

Fingerprint

Data transfer
Communication

ASJC Scopus subject areas

  • Software
  • Computer Science Applications

Cite this

Takizawa, H., Hirasawa, S., Sugawara, M., Gelado, I., Kobayashi, H., & Hwu, W-M. W. (2015). Optimized Data Transfers Based on the OpenCL Event Management Mechanism. Scientific Programming, 2015, [576498]. https://doi.org/10.1155/2015/576498

Optimized Data Transfers Based on the OpenCL Event Management Mechanism. / Takizawa, Hiroyuki; Hirasawa, Shoichi; Sugawara, Makoto; Gelado, Isaac; Kobayashi, Hiroaki; Hwu, Wen-Mei W.

In: Scientific Programming, Vol. 2015, 576498, 01.01.2015.

Research output: Contribution to journalArticle

Takizawa, H, Hirasawa, S, Sugawara, M, Gelado, I, Kobayashi, H & Hwu, W-MW 2015, 'Optimized Data Transfers Based on the OpenCL Event Management Mechanism', Scientific Programming, vol. 2015, 576498. https://doi.org/10.1155/2015/576498
Takizawa, Hiroyuki ; Hirasawa, Shoichi ; Sugawara, Makoto ; Gelado, Isaac ; Kobayashi, Hiroaki ; Hwu, Wen-Mei W. / Optimized Data Transfers Based on the OpenCL Event Management Mechanism. In: Scientific Programming. 2015 ; Vol. 2015.
@article{24cefd42ac7e443194ca69b191bbed37,
title = "Optimized Data Transfers Based on the OpenCL Event Management Mechanism",
abstract = "In standard OpenCL programming, hosts are supposed to control their compute devices. Since compute devices are dedicated to kernel computation, only hosts can execute several kinds of data transfers such as internode communication and file access. These data transfers require one host to simultaneously play two or more roles due to the need for collaboration between the host and devices. The codes for such data transfers are likely to be system-specific, resulting in low portability. This paper proposes an OpenCL extension that incorporates such data transfers into the OpenCL event management mechanism. Unlike the current OpenCL standard, the main thread running on the host is not blocked to serialize dependent operations. Hence, an application can easily use the opportunities to overlap parallel activities of hosts and compute devices. In addition, the implementation details of data transfers are hidden behind the extension, and application programmers can use the optimized data transfers without any tricky programming techniques. The evaluation results show that the proposed extension can use the optimized data transfer implementation and thereby increase the sustained data transfer performance by about 18{\%} for a real application accessing a big data file.",
author = "Hiroyuki Takizawa and Shoichi Hirasawa and Makoto Sugawara and Isaac Gelado and Hiroaki Kobayashi and Hwu, {Wen-Mei W}",
year = "2015",
month = "1",
day = "1",
doi = "10.1155/2015/576498",
language = "English (US)",
volume = "2015",
journal = "Scientific Programming",
issn = "1058-9244",
publisher = "IOS Press",

}

TY - JOUR

T1 - Optimized Data Transfers Based on the OpenCL Event Management Mechanism

AU - Takizawa, Hiroyuki

AU - Hirasawa, Shoichi

AU - Sugawara, Makoto

AU - Gelado, Isaac

AU - Kobayashi, Hiroaki

AU - Hwu, Wen-Mei W

PY - 2015/1/1

Y1 - 2015/1/1

N2 - In standard OpenCL programming, hosts are supposed to control their compute devices. Since compute devices are dedicated to kernel computation, only hosts can execute several kinds of data transfers such as internode communication and file access. These data transfers require one host to simultaneously play two or more roles due to the need for collaboration between the host and devices. The codes for such data transfers are likely to be system-specific, resulting in low portability. This paper proposes an OpenCL extension that incorporates such data transfers into the OpenCL event management mechanism. Unlike the current OpenCL standard, the main thread running on the host is not blocked to serialize dependent operations. Hence, an application can easily use the opportunities to overlap parallel activities of hosts and compute devices. In addition, the implementation details of data transfers are hidden behind the extension, and application programmers can use the optimized data transfers without any tricky programming techniques. The evaluation results show that the proposed extension can use the optimized data transfer implementation and thereby increase the sustained data transfer performance by about 18% for a real application accessing a big data file.

AB - In standard OpenCL programming, hosts are supposed to control their compute devices. Since compute devices are dedicated to kernel computation, only hosts can execute several kinds of data transfers such as internode communication and file access. These data transfers require one host to simultaneously play two or more roles due to the need for collaboration between the host and devices. The codes for such data transfers are likely to be system-specific, resulting in low portability. This paper proposes an OpenCL extension that incorporates such data transfers into the OpenCL event management mechanism. Unlike the current OpenCL standard, the main thread running on the host is not blocked to serialize dependent operations. Hence, an application can easily use the opportunities to overlap parallel activities of hosts and compute devices. In addition, the implementation details of data transfers are hidden behind the extension, and application programmers can use the optimized data transfers without any tricky programming techniques. The evaluation results show that the proposed extension can use the optimized data transfer implementation and thereby increase the sustained data transfer performance by about 18% for a real application accessing a big data file.

UR - http://www.scopus.com/inward/record.url?scp=84944224835&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84944224835&partnerID=8YFLogxK

U2 - 10.1155/2015/576498

DO - 10.1155/2015/576498

M3 - Article

AN - SCOPUS:84944224835

VL - 2015

JO - Scientific Programming

JF - Scientific Programming

SN - 1058-9244

M1 - 576498

ER -