CaCAO: Complex and compositional atomic operations for NoC-based manycore platforms

Sven Rheindt; Andreas Schenk; Akshay Srivatsa; Thomas Wild; Andreas Herkersdorf

doi:10.1007/978-3-319-77610-1_11

CaCAO: Complex and compositional atomic operations for NoC-based manycore platforms

Sven Rheindt, Andreas Schenk, Akshay Srivatsa, Thomas Wild, Andreas Herkersdorf

Chair of Integrated Systems

Technical University of Munich

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

5 Scopus citations

Abstract

Tile-based distributed memory systems have increased the scalability of manycore platforms. However, inter-tile memory accesses, especially thread synchronization suffer from high remote access latencies. Our thorough investigations of lock-based and lock-free synchronization primitives show that there is a concurrency dependent cross-over point between them, i.e. there is no one-fits-all solution. Therefore, we propose to combine the conceptual advantages (no retries and lock-free) of both variants by using dedicated hardware support for inter-tile atomic operations. For frequently used and highly concurrent data structures, we show a speedup factor of 23.9 and 35.4 over the lock-based and lock-free implementations respectively, which increases with higher concurrency.

Original language	English
Title of host publication	Architecture of Computing Systems – ARCS 2018 - 31st International Conference, Proceedings
Editors	Thilo Pionteck, Mladen Berekovic, Rainer Buchty, Dirk Koch, Heiko Hamann
Publisher	Springer Verlag
Pages	139-152
Number of pages	14
ISBN (Print)	9783319776095
DOIs	https://doi.org/10.1007/978-3-319-77610-1_11
State	Published - 2018
Event	31st International Conference on Architecture of Computing Systems, ARCS 2018 - Braunschweig, Germany Duration: 9 Apr 2018 → 12 Apr 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10793 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	31st International Conference on Architecture of Computing Systems, ARCS 2018
Country/Territory	Germany
City	Braunschweig
Period	9/04/18 → 12/04/18

Keywords

Atomic operations
Compare-and-swap
Distributed shared memory
Network-on-Chip
Remote synchronization

Access to Document

10.1007/978-3-319-77610-1_11

Cite this

Rheindt, S., Schenk, A., Srivatsa, A., Wild, T., & Herkersdorf, A. (2018). CaCAO: Complex and compositional atomic operations for NoC-based manycore platforms. In T. Pionteck, M. Berekovic, R. Buchty, D. Koch, & H. Hamann (Eds.), Architecture of Computing Systems – ARCS 2018 - 31st International Conference, Proceedings (pp. 139-152). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10793 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-77610-1_11

Rheindt, Sven ; Schenk, Andreas ; Srivatsa, Akshay et al. / CaCAO : Complex and compositional atomic operations for NoC-based manycore platforms. Architecture of Computing Systems – ARCS 2018 - 31st International Conference, Proceedings. editor / Thilo Pionteck ; Mladen Berekovic ; Rainer Buchty ; Dirk Koch ; Heiko Hamann. Springer Verlag, 2018. pp. 139-152 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{038204d306e74c458c07eeafed2d45cd,

title = "CaCAO: Complex and compositional atomic operations for NoC-based manycore platforms",

abstract = "Tile-based distributed memory systems have increased the scalability of manycore platforms. However, inter-tile memory accesses, especially thread synchronization suffer from high remote access latencies. Our thorough investigations of lock-based and lock-free synchronization primitives show that there is a concurrency dependent cross-over point between them, i.e. there is no one-fits-all solution. Therefore, we propose to combine the conceptual advantages (no retries and lock-free) of both variants by using dedicated hardware support for inter-tile atomic operations. For frequently used and highly concurrent data structures, we show a speedup factor of 23.9 and 35.4 over the lock-based and lock-free implementations respectively, which increases with higher concurrency.",

keywords = "Atomic operations, Compare-and-swap, Distributed shared memory, Network-on-Chip, Remote synchronization",

author = "Sven Rheindt and Andreas Schenk and Akshay Srivatsa and Thomas Wild and Andreas Herkersdorf",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing AG, part of Springer Nature 2018.; 31st International Conference on Architecture of Computing Systems, ARCS 2018 ; Conference date: 09-04-2018 Through 12-04-2018",

year = "2018",

doi = "10.1007/978-3-319-77610-1_11",

language = "English",

isbn = "9783319776095",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "139--152",

editor = "Thilo Pionteck and Mladen Berekovic and Rainer Buchty and Dirk Koch and Heiko Hamann",

booktitle = "Architecture of Computing Systems – ARCS 2018 - 31st International Conference, Proceedings",

}

Rheindt, S, Schenk, A, Srivatsa, A, Wild, T & Herkersdorf, A 2018, CaCAO: Complex and compositional atomic operations for NoC-based manycore platforms. in T Pionteck, M Berekovic, R Buchty, D Koch & H Hamann (eds), Architecture of Computing Systems – ARCS 2018 - 31st International Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10793 LNCS, Springer Verlag, pp. 139-152, 31st International Conference on Architecture of Computing Systems, ARCS 2018, Braunschweig, Germany, 9/04/18. https://doi.org/10.1007/978-3-319-77610-1_11

CaCAO: Complex and compositional atomic operations for NoC-based manycore platforms. / Rheindt, Sven; Schenk, Andreas; Srivatsa, Akshay et al.
Architecture of Computing Systems – ARCS 2018 - 31st International Conference, Proceedings. ed. / Thilo Pionteck; Mladen Berekovic; Rainer Buchty; Dirk Koch; Heiko Hamann. Springer Verlag, 2018. p. 139-152 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10793 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - CaCAO

T2 - 31st International Conference on Architecture of Computing Systems, ARCS 2018

AU - Rheindt, Sven

AU - Schenk, Andreas

AU - Srivatsa, Akshay

AU - Wild, Thomas

AU - Herkersdorf, Andreas

N1 - Publisher Copyright: © Springer International Publishing AG, part of Springer Nature 2018.

PY - 2018

Y1 - 2018

N2 - Tile-based distributed memory systems have increased the scalability of manycore platforms. However, inter-tile memory accesses, especially thread synchronization suffer from high remote access latencies. Our thorough investigations of lock-based and lock-free synchronization primitives show that there is a concurrency dependent cross-over point between them, i.e. there is no one-fits-all solution. Therefore, we propose to combine the conceptual advantages (no retries and lock-free) of both variants by using dedicated hardware support for inter-tile atomic operations. For frequently used and highly concurrent data structures, we show a speedup factor of 23.9 and 35.4 over the lock-based and lock-free implementations respectively, which increases with higher concurrency.

AB - Tile-based distributed memory systems have increased the scalability of manycore platforms. However, inter-tile memory accesses, especially thread synchronization suffer from high remote access latencies. Our thorough investigations of lock-based and lock-free synchronization primitives show that there is a concurrency dependent cross-over point between them, i.e. there is no one-fits-all solution. Therefore, we propose to combine the conceptual advantages (no retries and lock-free) of both variants by using dedicated hardware support for inter-tile atomic operations. For frequently used and highly concurrent data structures, we show a speedup factor of 23.9 and 35.4 over the lock-based and lock-free implementations respectively, which increases with higher concurrency.

KW - Atomic operations

KW - Compare-and-swap

KW - Distributed shared memory

KW - Network-on-Chip

KW - Remote synchronization

UR - http://www.scopus.com/inward/record.url?scp=85045459981&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-77610-1_11

DO - 10.1007/978-3-319-77610-1_11

M3 - Conference contribution

AN - SCOPUS:85045459981

SN - 9783319776095

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 139

EP - 152

BT - Architecture of Computing Systems – ARCS 2018 - 31st International Conference, Proceedings

A2 - Pionteck, Thilo

A2 - Berekovic, Mladen

A2 - Buchty, Rainer

A2 - Koch, Dirk

A2 - Hamann, Heiko

PB - Springer Verlag

Y2 - 9 April 2018 through 12 April 2018

ER -

Rheindt S, Schenk A, Srivatsa A, Wild T, Herkersdorf A. CaCAO: Complex and compositional atomic operations for NoC-based manycore platforms. In Pionteck T, Berekovic M, Buchty R, Koch D, Hamann H, editors, Architecture of Computing Systems – ARCS 2018 - 31st International Conference, Proceedings. Springer Verlag. 2018. p. 139-152. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-77610-1_11

CaCAO: Complex and compositional atomic operations for NoC-based manycore platforms

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this