In-network caching for chip multiprocessors

Aditya Yanamandra, Mary Jane Irwin, Vijaykrishnan Narayanan, Mahmut Kandemir, Sri Hari Krishna Narayanan

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

2 Citations (Scopus)

Abstract

Effective management of data is critical to the performance of emerging multi-core architectures. Our analysis of applications from SpecOMP reveal that a small fraction of shared addresses correspond to a large portion of accesses. Utilizing this observation, we propose a technique that augments a router in a on-chip network with a small data store to reduce the memory access latency of the shared data. In the proposed technique, shared data from read response packets that pass through the router are cached in its data store to reduce number of hops required to service future read requests. Our limit study reveals that such caching has the potential to reduce memory access latency on an average by 27%. Further, two practical caching strategies are shown to reduce memory access latency by 14% and 17% respectively with a data store of just four entries at 2.5% area overhead.

Original languageEnglish (US)
Title of host publicationHigh Performance Embedded Architectures and Compilers - Fourth International Conference, HiPEAC 2009, Proceedings
Pages373-388
Number of pages16
DOIs
StatePublished - Feb 4 2009
Event4th International Conference on High Performance Embedded Architectures and Compilers, HiPEAC 2009 - Paphos, Cyprus
Duration: Jan 25 2009Jan 28 2009

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume5409 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other4th International Conference on High Performance Embedded Architectures and Compilers, HiPEAC 2009
CountryCyprus
CityPaphos
Period1/25/091/28/09

Fingerprint

Chip multiprocessors
Caching
Routers
Data storage equipment
Latency
Router

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Yanamandra, A., Irwin, M. J., Narayanan, V., Kandemir, M., & Narayanan, S. H. K. (2009). In-network caching for chip multiprocessors. In High Performance Embedded Architectures and Compilers - Fourth International Conference, HiPEAC 2009, Proceedings (pp. 373-388). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5409 LNCS). https://doi.org/10.1007/978-3-540-92990-1_27
Yanamandra, Aditya ; Irwin, Mary Jane ; Narayanan, Vijaykrishnan ; Kandemir, Mahmut ; Narayanan, Sri Hari Krishna. / In-network caching for chip multiprocessors. High Performance Embedded Architectures and Compilers - Fourth International Conference, HiPEAC 2009, Proceedings. 2009. pp. 373-388 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{35981dcf92d64380aebf8a153dafe893,
title = "In-network caching for chip multiprocessors",
abstract = "Effective management of data is critical to the performance of emerging multi-core architectures. Our analysis of applications from SpecOMP reveal that a small fraction of shared addresses correspond to a large portion of accesses. Utilizing this observation, we propose a technique that augments a router in a on-chip network with a small data store to reduce the memory access latency of the shared data. In the proposed technique, shared data from read response packets that pass through the router are cached in its data store to reduce number of hops required to service future read requests. Our limit study reveals that such caching has the potential to reduce memory access latency on an average by 27{\%}. Further, two practical caching strategies are shown to reduce memory access latency by 14{\%} and 17{\%} respectively with a data store of just four entries at 2.5{\%} area overhead.",
author = "Aditya Yanamandra and Irwin, {Mary Jane} and Vijaykrishnan Narayanan and Mahmut Kandemir and Narayanan, {Sri Hari Krishna}",
year = "2009",
month = "2",
day = "4",
doi = "10.1007/978-3-540-92990-1_27",
language = "English (US)",
isbn = "3540929894",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "373--388",
booktitle = "High Performance Embedded Architectures and Compilers - Fourth International Conference, HiPEAC 2009, Proceedings",

}

Yanamandra, A, Irwin, MJ, Narayanan, V, Kandemir, M & Narayanan, SHK 2009, In-network caching for chip multiprocessors. in High Performance Embedded Architectures and Compilers - Fourth International Conference, HiPEAC 2009, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 5409 LNCS, pp. 373-388, 4th International Conference on High Performance Embedded Architectures and Compilers, HiPEAC 2009, Paphos, Cyprus, 1/25/09. https://doi.org/10.1007/978-3-540-92990-1_27

In-network caching for chip multiprocessors. / Yanamandra, Aditya; Irwin, Mary Jane; Narayanan, Vijaykrishnan; Kandemir, Mahmut; Narayanan, Sri Hari Krishna.

High Performance Embedded Architectures and Compilers - Fourth International Conference, HiPEAC 2009, Proceedings. 2009. p. 373-388 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5409 LNCS).

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

TY - GEN

T1 - In-network caching for chip multiprocessors

AU - Yanamandra, Aditya

AU - Irwin, Mary Jane

AU - Narayanan, Vijaykrishnan

AU - Kandemir, Mahmut

AU - Narayanan, Sri Hari Krishna

PY - 2009/2/4

Y1 - 2009/2/4

N2 - Effective management of data is critical to the performance of emerging multi-core architectures. Our analysis of applications from SpecOMP reveal that a small fraction of shared addresses correspond to a large portion of accesses. Utilizing this observation, we propose a technique that augments a router in a on-chip network with a small data store to reduce the memory access latency of the shared data. In the proposed technique, shared data from read response packets that pass through the router are cached in its data store to reduce number of hops required to service future read requests. Our limit study reveals that such caching has the potential to reduce memory access latency on an average by 27%. Further, two practical caching strategies are shown to reduce memory access latency by 14% and 17% respectively with a data store of just four entries at 2.5% area overhead.

AB - Effective management of data is critical to the performance of emerging multi-core architectures. Our analysis of applications from SpecOMP reveal that a small fraction of shared addresses correspond to a large portion of accesses. Utilizing this observation, we propose a technique that augments a router in a on-chip network with a small data store to reduce the memory access latency of the shared data. In the proposed technique, shared data from read response packets that pass through the router are cached in its data store to reduce number of hops required to service future read requests. Our limit study reveals that such caching has the potential to reduce memory access latency on an average by 27%. Further, two practical caching strategies are shown to reduce memory access latency by 14% and 17% respectively with a data store of just four entries at 2.5% area overhead.

UR - http://www.scopus.com/inward/record.url?scp=59049107362&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=59049107362&partnerID=8YFLogxK

U2 - 10.1007/978-3-540-92990-1_27

DO - 10.1007/978-3-540-92990-1_27

M3 - Conference contribution

SN - 3540929894

SN - 9783540929895

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

SP - 373

EP - 388

BT - High Performance Embedded Architectures and Compilers - Fourth International Conference, HiPEAC 2009, Proceedings

ER -

Yanamandra A, Irwin MJ, Narayanan V, Kandemir M, Narayanan SHK. In-network caching for chip multiprocessors. In High Performance Embedded Architectures and Compilers - Fourth International Conference, HiPEAC 2009, Proceedings. 2009. p. 373-388. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-540-92990-1_27