Power-Efficient Breadth-First Search with DRAM Row Buffer Locality-Aware Address Mapping

Satoshi Imamura, Yuichiro Yasui, Inoue Koji, Takatsugu Ono, Hiroshi Sasaki, Katsuki Fujisawa

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Citations (Scopus)

Abstract

Graph analysis applications have been widely used in real services such as road-traffic analysis and social network services. Breadth-first search (BFS) is one of the most representative algorithms for such applications; therefore, many researchers have tuned it to maximize performance. On the other hand, owing to the strict power constraints of modern HPC systems, it is necessary to improve power efficiency (i.e., performance per watt) when executing BFS. In this work, we focus on the power efficiency of DRAM and investigate the memory access pattern of a state-of-the-art BFS implementation using a cycle-accurate processor simulator. The results reveal that the conventional address mapping schemes of modern memory controllers do not efficiently exploit row buffers in DRAM. Thus, we propose a new scheme called per-row channel interleaving and improve the DRAM power efficiency by 30.3% compared to a conventional scheme for a certain simulator setting. Moreover, we demonstrate that this proposed scheme is effective for various configurations of memory controllers.

Original languageEnglish
Title of host publicationProceedings of HPGDMP 2016
Subtitle of host publicationHigh Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages17-24
Number of pages8
ISBN (Electronic)9781509038800
DOIs
Publication statusPublished - Jan 23 2017
Event2016 High Performance Graph Data Management and Processing, HPGDMP 2016 - Salt Lake City, United States
Duration: Nov 13 2016 → …

Publication series

NameProceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis

Other

Other2016 High Performance Graph Data Management and Processing, HPGDMP 2016
CountryUnited States
CitySalt Lake City
Period11/13/16 → …

Fingerprint

Dynamic random access storage
Data storage equipment
Simulators
Controllers

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture

Cite this

Imamura, S., Yasui, Y., Koji, I., Ono, T., Sasaki, H., & Fujisawa, K. (2017). Power-Efficient Breadth-First Search with DRAM Row Buffer Locality-Aware Address Mapping. In Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis (pp. 17-24). [7830443] (Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/HPGDMP.2016.010

Power-Efficient Breadth-First Search with DRAM Row Buffer Locality-Aware Address Mapping. / Imamura, Satoshi; Yasui, Yuichiro; Koji, Inoue; Ono, Takatsugu; Sasaki, Hiroshi; Fujisawa, Katsuki.

Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers Inc., 2017. p. 17-24 7830443 (Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Imamura, S, Yasui, Y, Koji, I, Ono, T, Sasaki, H & Fujisawa, K 2017, Power-Efficient Breadth-First Search with DRAM Row Buffer Locality-Aware Address Mapping. in Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis., 7830443, Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis, Institute of Electrical and Electronics Engineers Inc., pp. 17-24, 2016 High Performance Graph Data Management and Processing, HPGDMP 2016, Salt Lake City, United States, 11/13/16. https://doi.org/10.1109/HPGDMP.2016.010
Imamura S, Yasui Y, Koji I, Ono T, Sasaki H, Fujisawa K. Power-Efficient Breadth-First Search with DRAM Row Buffer Locality-Aware Address Mapping. In Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers Inc. 2017. p. 17-24. 7830443. (Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis). https://doi.org/10.1109/HPGDMP.2016.010
Imamura, Satoshi ; Yasui, Yuichiro ; Koji, Inoue ; Ono, Takatsugu ; Sasaki, Hiroshi ; Fujisawa, Katsuki. / Power-Efficient Breadth-First Search with DRAM Row Buffer Locality-Aware Address Mapping. Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis. Institute of Electrical and Electronics Engineers Inc., 2017. pp. 17-24 (Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis).
@inproceedings{ec2896b20e6042e4be605679dad7684e,
title = "Power-Efficient Breadth-First Search with DRAM Row Buffer Locality-Aware Address Mapping",
abstract = "Graph analysis applications have been widely used in real services such as road-traffic analysis and social network services. Breadth-first search (BFS) is one of the most representative algorithms for such applications; therefore, many researchers have tuned it to maximize performance. On the other hand, owing to the strict power constraints of modern HPC systems, it is necessary to improve power efficiency (i.e., performance per watt) when executing BFS. In this work, we focus on the power efficiency of DRAM and investigate the memory access pattern of a state-of-the-art BFS implementation using a cycle-accurate processor simulator. The results reveal that the conventional address mapping schemes of modern memory controllers do not efficiently exploit row buffers in DRAM. Thus, we propose a new scheme called per-row channel interleaving and improve the DRAM power efficiency by 30.3{\%} compared to a conventional scheme for a certain simulator setting. Moreover, we demonstrate that this proposed scheme is effective for various configurations of memory controllers.",
author = "Satoshi Imamura and Yuichiro Yasui and Inoue Koji and Takatsugu Ono and Hiroshi Sasaki and Katsuki Fujisawa",
year = "2017",
month = "1",
day = "23",
doi = "10.1109/HPGDMP.2016.010",
language = "English",
series = "Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "17--24",
booktitle = "Proceedings of HPGDMP 2016",
address = "United States",

}

TY - GEN

T1 - Power-Efficient Breadth-First Search with DRAM Row Buffer Locality-Aware Address Mapping

AU - Imamura, Satoshi

AU - Yasui, Yuichiro

AU - Koji, Inoue

AU - Ono, Takatsugu

AU - Sasaki, Hiroshi

AU - Fujisawa, Katsuki

PY - 2017/1/23

Y1 - 2017/1/23

N2 - Graph analysis applications have been widely used in real services such as road-traffic analysis and social network services. Breadth-first search (BFS) is one of the most representative algorithms for such applications; therefore, many researchers have tuned it to maximize performance. On the other hand, owing to the strict power constraints of modern HPC systems, it is necessary to improve power efficiency (i.e., performance per watt) when executing BFS. In this work, we focus on the power efficiency of DRAM and investigate the memory access pattern of a state-of-the-art BFS implementation using a cycle-accurate processor simulator. The results reveal that the conventional address mapping schemes of modern memory controllers do not efficiently exploit row buffers in DRAM. Thus, we propose a new scheme called per-row channel interleaving and improve the DRAM power efficiency by 30.3% compared to a conventional scheme for a certain simulator setting. Moreover, we demonstrate that this proposed scheme is effective for various configurations of memory controllers.

AB - Graph analysis applications have been widely used in real services such as road-traffic analysis and social network services. Breadth-first search (BFS) is one of the most representative algorithms for such applications; therefore, many researchers have tuned it to maximize performance. On the other hand, owing to the strict power constraints of modern HPC systems, it is necessary to improve power efficiency (i.e., performance per watt) when executing BFS. In this work, we focus on the power efficiency of DRAM and investigate the memory access pattern of a state-of-the-art BFS implementation using a cycle-accurate processor simulator. The results reveal that the conventional address mapping schemes of modern memory controllers do not efficiently exploit row buffers in DRAM. Thus, we propose a new scheme called per-row channel interleaving and improve the DRAM power efficiency by 30.3% compared to a conventional scheme for a certain simulator setting. Moreover, we demonstrate that this proposed scheme is effective for various configurations of memory controllers.

UR - http://www.scopus.com/inward/record.url?scp=85013899496&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85013899496&partnerID=8YFLogxK

U2 - 10.1109/HPGDMP.2016.010

DO - 10.1109/HPGDMP.2016.010

M3 - Conference contribution

T3 - Proceedings of HPGDMP 2016: High Performance Graph Data Management and Processing - Held in conjunction with SC 2016: The International Conference for High Performance Computing, Networking, Storage and Analysis

SP - 17

EP - 24

BT - Proceedings of HPGDMP 2016

PB - Institute of Electrical and Electronics Engineers Inc.

ER -