RING-DATA ORDER: A new cache coherence protocol for ring-based multicores

Jin Young Park, Lynn Choi

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

Abstract

RING-DATA ORDER mechanism is motivated to solve cache coherence ordering for ring-based multicores. For point to point ring interconnects, the existing ORDERING-POINT mechanism is known for poor average performance due to its long response latency. GREEDY-ORDER mechanism is simple but the performance is still unacceptable because the requestor should often retry to complete a coherent operation. RING-ORDER mechanism is the best known algorithm so far but special storage and management overhead are expected due to its token management. This paper proposes a new mechanism called RING-DATA ORDER which is as simple as GREEDY-ORDER but more efficient than RING-ORDER. RING-DATA ORDER determines the order by data transfer sequence. The unbounded retries of GREEDY-ORDER can be eliminated by blocking incoming coherence request which contains data. Blocked request restarts to traverse the ring interconnect when the coherence operation for current node is completed. Blocking is also used in RINGORDER mechanism. But RING-DATA ORDER also eliminates token and its additional management overhead without sacrificing performance. Using a custom-built multicore simulator with profiled SPEC 2000 integer benchmark suites we demonstrate that RING-DATA ORDER can achieve the same level of performance as RING-ORDER without the overhead of token management.

Original languageEnglish
Title of host publicationProceedings of the 2009 International Conference on High Performance Computing and Simulation, HPCS 2009
Pages82-88
Number of pages7
DOIs
Publication statusPublished - 2009 Nov 19
Event2009 International Conference on High Performance Computing and Simulation, HPCS '09 - Leipzig, Germany
Duration: 2009 Jun 212009 Jun 24

Other

Other2009 International Conference on High Performance Computing and Simulation, HPCS '09
CountryGermany
CityLeipzig
Period09/6/2109/6/24

Fingerprint

Data transfer
Simulators

Keywords

  • Cache
  • Coherence
  • Multicore
  • Ordering

ASJC Scopus subject areas

  • Computer Science(all)

Cite this

Park, J. Y., & Choi, L. (2009). RING-DATA ORDER: A new cache coherence protocol for ring-based multicores. In Proceedings of the 2009 International Conference on High Performance Computing and Simulation, HPCS 2009 (pp. 82-88). [5192668] https://doi.org/10.1109/HPCSIM.2009.5192668

RING-DATA ORDER : A new cache coherence protocol for ring-based multicores. / Park, Jin Young; Choi, Lynn.

Proceedings of the 2009 International Conference on High Performance Computing and Simulation, HPCS 2009. 2009. p. 82-88 5192668.

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

Park, JY & Choi, L 2009, RING-DATA ORDER: A new cache coherence protocol for ring-based multicores. in Proceedings of the 2009 International Conference on High Performance Computing and Simulation, HPCS 2009., 5192668, pp. 82-88, 2009 International Conference on High Performance Computing and Simulation, HPCS '09, Leipzig, Germany, 09/6/21. https://doi.org/10.1109/HPCSIM.2009.5192668
Park JY, Choi L. RING-DATA ORDER: A new cache coherence protocol for ring-based multicores. In Proceedings of the 2009 International Conference on High Performance Computing and Simulation, HPCS 2009. 2009. p. 82-88. 5192668 https://doi.org/10.1109/HPCSIM.2009.5192668
Park, Jin Young ; Choi, Lynn. / RING-DATA ORDER : A new cache coherence protocol for ring-based multicores. Proceedings of the 2009 International Conference on High Performance Computing and Simulation, HPCS 2009. 2009. pp. 82-88
@inproceedings{9774219e4df34c6eb5d8bc9728bc3e14,
title = "RING-DATA ORDER: A new cache coherence protocol for ring-based multicores",
abstract = "RING-DATA ORDER mechanism is motivated to solve cache coherence ordering for ring-based multicores. For point to point ring interconnects, the existing ORDERING-POINT mechanism is known for poor average performance due to its long response latency. GREEDY-ORDER mechanism is simple but the performance is still unacceptable because the requestor should often retry to complete a coherent operation. RING-ORDER mechanism is the best known algorithm so far but special storage and management overhead are expected due to its token management. This paper proposes a new mechanism called RING-DATA ORDER which is as simple as GREEDY-ORDER but more efficient than RING-ORDER. RING-DATA ORDER determines the order by data transfer sequence. The unbounded retries of GREEDY-ORDER can be eliminated by blocking incoming coherence request which contains data. Blocked request restarts to traverse the ring interconnect when the coherence operation for current node is completed. Blocking is also used in RINGORDER mechanism. But RING-DATA ORDER also eliminates token and its additional management overhead without sacrificing performance. Using a custom-built multicore simulator with profiled SPEC 2000 integer benchmark suites we demonstrate that RING-DATA ORDER can achieve the same level of performance as RING-ORDER without the overhead of token management.",
keywords = "Cache, Coherence, Multicore, Ordering",
author = "Park, {Jin Young} and Lynn Choi",
year = "2009",
month = "11",
day = "19",
doi = "10.1109/HPCSIM.2009.5192668",
language = "English",
isbn = "9781424449071",
pages = "82--88",
booktitle = "Proceedings of the 2009 International Conference on High Performance Computing and Simulation, HPCS 2009",

}

TY - GEN

T1 - RING-DATA ORDER

T2 - A new cache coherence protocol for ring-based multicores

AU - Park, Jin Young

AU - Choi, Lynn

PY - 2009/11/19

Y1 - 2009/11/19

N2 - RING-DATA ORDER mechanism is motivated to solve cache coherence ordering for ring-based multicores. For point to point ring interconnects, the existing ORDERING-POINT mechanism is known for poor average performance due to its long response latency. GREEDY-ORDER mechanism is simple but the performance is still unacceptable because the requestor should often retry to complete a coherent operation. RING-ORDER mechanism is the best known algorithm so far but special storage and management overhead are expected due to its token management. This paper proposes a new mechanism called RING-DATA ORDER which is as simple as GREEDY-ORDER but more efficient than RING-ORDER. RING-DATA ORDER determines the order by data transfer sequence. The unbounded retries of GREEDY-ORDER can be eliminated by blocking incoming coherence request which contains data. Blocked request restarts to traverse the ring interconnect when the coherence operation for current node is completed. Blocking is also used in RINGORDER mechanism. But RING-DATA ORDER also eliminates token and its additional management overhead without sacrificing performance. Using a custom-built multicore simulator with profiled SPEC 2000 integer benchmark suites we demonstrate that RING-DATA ORDER can achieve the same level of performance as RING-ORDER without the overhead of token management.

AB - RING-DATA ORDER mechanism is motivated to solve cache coherence ordering for ring-based multicores. For point to point ring interconnects, the existing ORDERING-POINT mechanism is known for poor average performance due to its long response latency. GREEDY-ORDER mechanism is simple but the performance is still unacceptable because the requestor should often retry to complete a coherent operation. RING-ORDER mechanism is the best known algorithm so far but special storage and management overhead are expected due to its token management. This paper proposes a new mechanism called RING-DATA ORDER which is as simple as GREEDY-ORDER but more efficient than RING-ORDER. RING-DATA ORDER determines the order by data transfer sequence. The unbounded retries of GREEDY-ORDER can be eliminated by blocking incoming coherence request which contains data. Blocked request restarts to traverse the ring interconnect when the coherence operation for current node is completed. Blocking is also used in RINGORDER mechanism. But RING-DATA ORDER also eliminates token and its additional management overhead without sacrificing performance. Using a custom-built multicore simulator with profiled SPEC 2000 integer benchmark suites we demonstrate that RING-DATA ORDER can achieve the same level of performance as RING-ORDER without the overhead of token management.

KW - Cache

KW - Coherence

KW - Multicore

KW - Ordering

UR - http://www.scopus.com/inward/record.url?scp=70449491669&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70449491669&partnerID=8YFLogxK

U2 - 10.1109/HPCSIM.2009.5192668

DO - 10.1109/HPCSIM.2009.5192668

M3 - Conference contribution

AN - SCOPUS:70449491669

SN - 9781424449071

SP - 82

EP - 88

BT - Proceedings of the 2009 International Conference on High Performance Computing and Simulation, HPCS 2009

ER -