Supporting concurrent memory access and multioperations in moving threads CMPs

Martti Forsell, Ville Leppänen

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

Abstract

Moving threads is a theoretically interesting approach for mapping the computation of an application to chip multiprocessor (CMP) architectures. Instead of moving data read and write requests, extremely lightweight threads are moved between the processor cores. As a consequence, all kinds of cache coherence problems and need for read reply messages are eliminated. Although moving threads architectures in many ways support the same techniques as their moving data counterparts, the existing architectures support exclusive memory access only degrading the performance of the approach by a logarithmic factor in many algorithms with respect to the best moving data architectures. In this paper we propose an architectural technique supporting partial concurrent memory access and multioperations for a class of synchronous moving threads CMPs. Application examples and performance, silicon area, and power consumption evaluations are given.
Original languageEnglish
Title of host publicationProceedings of the 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, PDPTA 2010
PublisherCSREA Press
Pages377-383
ISBN (Print)1-60132-158-9
Publication statusPublished - 2010
MoE publication typeA4 Article in a conference publication
Event2010 International Conference on Parallel and Distributed Processing Techniques and Applications, PDPTA’10 - Las Vegas, Nevada, United States
Duration: 12 Jul 201015 Jul 2010

Conference

Conference2010 International Conference on Parallel and Distributed Processing Techniques and Applications, PDPTA’10
Abbreviated titlePDPTA 2010
CountryUnited States
CityLas Vegas, Nevada
Period12/07/1015/07/10

Fingerprint

Data storage equipment
Electric power utilization
Silicon

Keywords

  • Moving threads
  • parallel computing
  • CMP
  • concurrent memory access
  • computer architecture

Cite this

Forsell, M., & Leppänen, V. (2010). Supporting concurrent memory access and multioperations in moving threads CMPs. In Proceedings of the 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, PDPTA 2010 (pp. 377-383). CSREA Press.
Forsell, Martti ; Leppänen, Ville. / Supporting concurrent memory access and multioperations in moving threads CMPs. Proceedings of the 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, PDPTA 2010. CSREA Press, 2010. pp. 377-383
@inproceedings{c78a93c927fb457fbffe620cbb14360f,
title = "Supporting concurrent memory access and multioperations in moving threads CMPs",
abstract = "Moving threads is a theoretically interesting approach for mapping the computation of an application to chip multiprocessor (CMP) architectures. Instead of moving data read and write requests, extremely lightweight threads are moved between the processor cores. As a consequence, all kinds of cache coherence problems and need for read reply messages are eliminated. Although moving threads architectures in many ways support the same techniques as their moving data counterparts, the existing architectures support exclusive memory access only degrading the performance of the approach by a logarithmic factor in many algorithms with respect to the best moving data architectures. In this paper we propose an architectural technique supporting partial concurrent memory access and multioperations for a class of synchronous moving threads CMPs. Application examples and performance, silicon area, and power consumption evaluations are given.",
keywords = "Moving threads, parallel computing, CMP, concurrent memory access, computer architecture",
author = "Martti Forsell and Ville Lepp{\"a}nen",
year = "2010",
language = "English",
isbn = "1-60132-158-9",
pages = "377--383",
booktitle = "Proceedings of the 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, PDPTA 2010",
publisher = "CSREA Press",
address = "United States",

}

Forsell, M & Leppänen, V 2010, Supporting concurrent memory access and multioperations in moving threads CMPs. in Proceedings of the 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, PDPTA 2010. CSREA Press, pp. 377-383, 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, PDPTA’10, Las Vegas, Nevada, United States, 12/07/10.

Supporting concurrent memory access and multioperations in moving threads CMPs. / Forsell, Martti; Leppänen, Ville.

Proceedings of the 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, PDPTA 2010. CSREA Press, 2010. p. 377-383.

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

TY - GEN

T1 - Supporting concurrent memory access and multioperations in moving threads CMPs

AU - Forsell, Martti

AU - Leppänen, Ville

PY - 2010

Y1 - 2010

N2 - Moving threads is a theoretically interesting approach for mapping the computation of an application to chip multiprocessor (CMP) architectures. Instead of moving data read and write requests, extremely lightweight threads are moved between the processor cores. As a consequence, all kinds of cache coherence problems and need for read reply messages are eliminated. Although moving threads architectures in many ways support the same techniques as their moving data counterparts, the existing architectures support exclusive memory access only degrading the performance of the approach by a logarithmic factor in many algorithms with respect to the best moving data architectures. In this paper we propose an architectural technique supporting partial concurrent memory access and multioperations for a class of synchronous moving threads CMPs. Application examples and performance, silicon area, and power consumption evaluations are given.

AB - Moving threads is a theoretically interesting approach for mapping the computation of an application to chip multiprocessor (CMP) architectures. Instead of moving data read and write requests, extremely lightweight threads are moved between the processor cores. As a consequence, all kinds of cache coherence problems and need for read reply messages are eliminated. Although moving threads architectures in many ways support the same techniques as their moving data counterparts, the existing architectures support exclusive memory access only degrading the performance of the approach by a logarithmic factor in many algorithms with respect to the best moving data architectures. In this paper we propose an architectural technique supporting partial concurrent memory access and multioperations for a class of synchronous moving threads CMPs. Application examples and performance, silicon area, and power consumption evaluations are given.

KW - Moving threads

KW - parallel computing

KW - CMP

KW - concurrent memory access

KW - computer architecture

M3 - Conference article in proceedings

SN - 1-60132-158-9

SP - 377

EP - 383

BT - Proceedings of the 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, PDPTA 2010

PB - CSREA Press

ER -

Forsell M, Leppänen V. Supporting concurrent memory access and multioperations in moving threads CMPs. In Proceedings of the 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, PDPTA 2010. CSREA Press. 2010. p. 377-383