Primal-dual method based simultaneous functional unit and register binding

Jianmo Ni, Cong Hao, Nan Wang, Qian Ai, Takeshi Yoshimura

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

Abstract

Interconnect reduction is one of the key issues in high-level synthesis. In this paper, we propose a primal-dual based method to solve the functional unit (FU) and register binding simultaneously while minimizing the global interconnection. Specifically, the binding problem is formulated as a min-cost network flow based on splitting weighted and order compatibility graphs (SWOCGs). The interconnect sharing among registers and FUs are maximized by binding the the operations or variables on the same path to the same FUs or registers according to the flow. Experimental results show that, compared with the previous greedy method [7], our proposed algorithm achieves an average 4.8% further reduction in global interconnection for a suite of benchmarks.

Original languageEnglish
Title of host publicationProceedings - 2015 IEEE 11th International Conference on ASIC, ASICON 2015
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781479984831
DOIs
Publication statusPublished - 2016 Jul 19
Event11th IEEE International Conference on Advanced Semiconductor Integrated Circuits (ASIC), ASICON 2015 - Chengdu, China
Duration: 2015 Nov 32015 Nov 6

Other

Other11th IEEE International Conference on Advanced Semiconductor Integrated Circuits (ASIC), ASICON 2015
CountryChina
CityChengdu
Period15/11/315/11/6

Fingerprint

Costs
High level synthesis

ASJC Scopus subject areas

  • Hardware and Architecture
  • Electrical and Electronic Engineering

Cite this

Ni, J., Hao, C., Wang, N., Ai, Q., & Yoshimura, T. (2016). Primal-dual method based simultaneous functional unit and register binding. In Proceedings - 2015 IEEE 11th International Conference on ASIC, ASICON 2015 [7516909] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ASICON.2015.7516909

Primal-dual method based simultaneous functional unit and register binding. / Ni, Jianmo; Hao, Cong; Wang, Nan; Ai, Qian; Yoshimura, Takeshi.

Proceedings - 2015 IEEE 11th International Conference on ASIC, ASICON 2015. Institute of Electrical and Electronics Engineers Inc., 2016. 7516909.

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

Ni, J, Hao, C, Wang, N, Ai, Q & Yoshimura, T 2016, Primal-dual method based simultaneous functional unit and register binding. in Proceedings - 2015 IEEE 11th International Conference on ASIC, ASICON 2015., 7516909, Institute of Electrical and Electronics Engineers Inc., 11th IEEE International Conference on Advanced Semiconductor Integrated Circuits (ASIC), ASICON 2015, Chengdu, China, 15/11/3. https://doi.org/10.1109/ASICON.2015.7516909
Ni J, Hao C, Wang N, Ai Q, Yoshimura T. Primal-dual method based simultaneous functional unit and register binding. In Proceedings - 2015 IEEE 11th International Conference on ASIC, ASICON 2015. Institute of Electrical and Electronics Engineers Inc. 2016. 7516909 https://doi.org/10.1109/ASICON.2015.7516909
Ni, Jianmo ; Hao, Cong ; Wang, Nan ; Ai, Qian ; Yoshimura, Takeshi. / Primal-dual method based simultaneous functional unit and register binding. Proceedings - 2015 IEEE 11th International Conference on ASIC, ASICON 2015. Institute of Electrical and Electronics Engineers Inc., 2016.
@inproceedings{442bb0f6a6864ea6b661563b6687d1ad,
title = "Primal-dual method based simultaneous functional unit and register binding",
abstract = "Interconnect reduction is one of the key issues in high-level synthesis. In this paper, we propose a primal-dual based method to solve the functional unit (FU) and register binding simultaneously while minimizing the global interconnection. Specifically, the binding problem is formulated as a min-cost network flow based on splitting weighted and order compatibility graphs (SWOCGs). The interconnect sharing among registers and FUs are maximized by binding the the operations or variables on the same path to the same FUs or registers according to the flow. Experimental results show that, compared with the previous greedy method [7], our proposed algorithm achieves an average 4.8{\%} further reduction in global interconnection for a suite of benchmarks.",
author = "Jianmo Ni and Cong Hao and Nan Wang and Qian Ai and Takeshi Yoshimura",
year = "2016",
month = "7",
day = "19",
doi = "10.1109/ASICON.2015.7516909",
language = "English",
booktitle = "Proceedings - 2015 IEEE 11th International Conference on ASIC, ASICON 2015",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
address = "United States",

}

TY - GEN

T1 - Primal-dual method based simultaneous functional unit and register binding

AU - Ni, Jianmo

AU - Hao, Cong

AU - Wang, Nan

AU - Ai, Qian

AU - Yoshimura, Takeshi

PY - 2016/7/19

Y1 - 2016/7/19

N2 - Interconnect reduction is one of the key issues in high-level synthesis. In this paper, we propose a primal-dual based method to solve the functional unit (FU) and register binding simultaneously while minimizing the global interconnection. Specifically, the binding problem is formulated as a min-cost network flow based on splitting weighted and order compatibility graphs (SWOCGs). The interconnect sharing among registers and FUs are maximized by binding the the operations or variables on the same path to the same FUs or registers according to the flow. Experimental results show that, compared with the previous greedy method [7], our proposed algorithm achieves an average 4.8% further reduction in global interconnection for a suite of benchmarks.

AB - Interconnect reduction is one of the key issues in high-level synthesis. In this paper, we propose a primal-dual based method to solve the functional unit (FU) and register binding simultaneously while minimizing the global interconnection. Specifically, the binding problem is formulated as a min-cost network flow based on splitting weighted and order compatibility graphs (SWOCGs). The interconnect sharing among registers and FUs are maximized by binding the the operations or variables on the same path to the same FUs or registers according to the flow. Experimental results show that, compared with the previous greedy method [7], our proposed algorithm achieves an average 4.8% further reduction in global interconnection for a suite of benchmarks.

UR - http://www.scopus.com/inward/record.url?scp=84982274177&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84982274177&partnerID=8YFLogxK

U2 - 10.1109/ASICON.2015.7516909

DO - 10.1109/ASICON.2015.7516909

M3 - Conference contribution

BT - Proceedings - 2015 IEEE 11th International Conference on ASIC, ASICON 2015

PB - Institute of Electrical and Electronics Engineers Inc.

ER -