NMPO: Near-Memory Computing Profiling and Offloading

Stefano Corda; Madhurya Kumaraswamy; Ahsan Javed Awan; Roel Jordans; Akash Kumar; Henk Corporaal

doi:10.1109/DSD53832.2021.00048

NMPO: Near-Memory Computing Profiling and Offloading

Stefano Corda, Madhurya Kumaraswamy, Ahsan Javed Awan, Roel Jordans, Akash Kumar, Henk Corporaal

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

8 Citations (Scopus)

106 Downloads (Pure)

Abstract

Real-world applications are now processing big-data sets, often bottlenecked by the data movement between the compute units and the main memory. Near-memory computing (NMC), a modern data-centric computational paradigm, can alleviate these bottlenecks, thereby improving the performance of applications. The lack of NMC system availability makes simulators the primary evaluation tool for performance estimation. However, simulators are usually time-consuming, and methods that can reduce this overhead would accelerate the early-stage design process of NMC systems. This work proposes Near-Memory computing Profiling and Offloading (NMPO), a high-level framework capable of predicting NMC offloading suitability employing an ensemble machine learning model. NMPO predicts NMC suitability with an accuracy of 85.6% and, compared to prior works, can reduce the prediction time by using hardware-dependent applications features by up to 3 order of magnitude.

Original language	English
Title of host publication	Proceedings - 2021 24th Euromicro Conference on Digital System Design, DSD 2021
Editors	Francesco Leporati, Salvatore Vitabile, Amund Skavhaug
Publisher	Institute of Electrical and Electronics Engineers
Pages	259-267
Number of pages	9
ISBN (Electronic)	978-1-6654-2703-6
DOIs	https://doi.org/10.1109/DSD53832.2021.00048
Publication status	Published - 11 Oct 2021
Event	24th Euromicro Conference on Digital System Design, DSD 2021 - Virtual, Online, Palermo, Italy Duration: 1 Sept 2021 → 3 Sept 2021 Conference number: 24

Conference

Conference	24th Euromicro Conference on Digital System Design, DSD 2021
Abbreviated title	DSD 2021
Country/Territory	Italy
City	Palermo
Period	1/09/21 → 3/09/21

Bibliographical note

Euromicro Conference on Digital System Design 2021

Funding

ACKNOWLEDGMENTS This work is funded by the European Commission under Marie Sklodowska-Curie Innovative Training Networks European Industrial Doctorate (Project ID: 676240). We would like to thank Gabor Nemeth from Ericsson Research for his feedback on the draft of the paper.

Funders	Funder number
European Union's Horizon 2020 - Research and Innovation Framework Programme
European Commission	676240

Keywords

cs.AR
cs.PF

Access to Document

10.1109/DSD53832.2021.00048

ArXiv 2106.15284v1Submitted manuscript, 755 KB

Cite this

Corda, S., Kumaraswamy, M., Awan, A. J., Jordans, R., Kumar, A., & Corporaal, H. (2021). NMPO: Near-Memory Computing Profiling and Offloading. In F. Leporati, S. Vitabile, & A. Skavhaug (Eds.), Proceedings - 2021 24th Euromicro Conference on Digital System Design, DSD 2021 (pp. 259-267). Article 9556449 Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/DSD53832.2021.00048

@inproceedings{98888edf38b74cedac7e5a1dd1ebe787,

title = "NMPO: Near-Memory Computing Profiling and Offloading",

abstract = "Real-world applications are now processing big-data sets, often bottlenecked by the data movement between the compute units and the main memory. Near-memory computing (NMC), a modern data-centric computational paradigm, can alleviate these bottlenecks, thereby improving the performance of applications. The lack of NMC system availability makes simulators the primary evaluation tool for performance estimation. However, simulators are usually time-consuming, and methods that can reduce this overhead would accelerate the early-stage design process of NMC systems. This work proposes Near-Memory computing Profiling and Offloading (NMPO), a high-level framework capable of predicting NMC offloading suitability employing an ensemble machine learning model. NMPO predicts NMC suitability with an accuracy of 85.6\% and, compared to prior works, can reduce the prediction time by using hardware-dependent applications features by up to 3 order of magnitude. ",

keywords = "cs.AR, cs.PF",

author = "Stefano Corda and Madhurya Kumaraswamy and Awan, \{Ahsan Javed\} and Roel Jordans and Akash Kumar and Henk Corporaal",

note = "Euromicro Conference on Digital System Design 2021; 24th Euromicro Conference on Digital System Design, DSD 2021, DSD 2021 ; Conference date: 01-09-2021 Through 03-09-2021",

year = "2021",

month = oct,

day = "11",

doi = "10.1109/DSD53832.2021.00048",

language = "English",

pages = "259--267",

editor = "Francesco Leporati and Salvatore Vitabile and Amund Skavhaug",

booktitle = "Proceedings - 2021 24th Euromicro Conference on Digital System Design, DSD 2021",

publisher = "Institute of Electrical and Electronics Engineers",

address = "United States",

}

Corda, S, Kumaraswamy, M, Awan, AJ, Jordans, R, Kumar, A & Corporaal, H 2021, NMPO: Near-Memory Computing Profiling and Offloading. in F Leporati, S Vitabile & A Skavhaug (eds), Proceedings - 2021 24th Euromicro Conference on Digital System Design, DSD 2021., 9556449, Institute of Electrical and Electronics Engineers, pp. 259-267, 24th Euromicro Conference on Digital System Design, DSD 2021, Palermo, Italy, 1/09/21. https://doi.org/10.1109/DSD53832.2021.00048

NMPO: Near-Memory Computing Profiling and Offloading. / Corda, Stefano; Kumaraswamy, Madhurya; Awan, Ahsan Javed et al.
Proceedings - 2021 24th Euromicro Conference on Digital System Design, DSD 2021. ed. / Francesco Leporati; Salvatore Vitabile; Amund Skavhaug. Institute of Electrical and Electronics Engineers, 2021. p. 259-267 9556449.

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

TY - GEN

T1 - NMPO

T2 - 24th Euromicro Conference on Digital System Design, DSD 2021

AU - Corda, Stefano

AU - Kumaraswamy, Madhurya

AU - Awan, Ahsan Javed

AU - Jordans, Roel

AU - Kumar, Akash

AU - Corporaal, Henk

N1 - Conference code: 24

PY - 2021/10/11

Y1 - 2021/10/11

N2 - Real-world applications are now processing big-data sets, often bottlenecked by the data movement between the compute units and the main memory. Near-memory computing (NMC), a modern data-centric computational paradigm, can alleviate these bottlenecks, thereby improving the performance of applications. The lack of NMC system availability makes simulators the primary evaluation tool for performance estimation. However, simulators are usually time-consuming, and methods that can reduce this overhead would accelerate the early-stage design process of NMC systems. This work proposes Near-Memory computing Profiling and Offloading (NMPO), a high-level framework capable of predicting NMC offloading suitability employing an ensemble machine learning model. NMPO predicts NMC suitability with an accuracy of 85.6% and, compared to prior works, can reduce the prediction time by using hardware-dependent applications features by up to 3 order of magnitude.

AB - Real-world applications are now processing big-data sets, often bottlenecked by the data movement between the compute units and the main memory. Near-memory computing (NMC), a modern data-centric computational paradigm, can alleviate these bottlenecks, thereby improving the performance of applications. The lack of NMC system availability makes simulators the primary evaluation tool for performance estimation. However, simulators are usually time-consuming, and methods that can reduce this overhead would accelerate the early-stage design process of NMC systems. This work proposes Near-Memory computing Profiling and Offloading (NMPO), a high-level framework capable of predicting NMC offloading suitability employing an ensemble machine learning model. NMPO predicts NMC suitability with an accuracy of 85.6% and, compared to prior works, can reduce the prediction time by using hardware-dependent applications features by up to 3 order of magnitude.

KW - cs.AR

KW - cs.PF

UR - http://www.scopus.com/inward/record.url?scp=85125803813&partnerID=8YFLogxK

U2 - 10.1109/DSD53832.2021.00048

DO - 10.1109/DSD53832.2021.00048

M3 - Conference contribution

SP - 259

EP - 267

BT - Proceedings - 2021 24th Euromicro Conference on Digital System Design, DSD 2021

A2 - Leporati, Francesco

A2 - Vitabile, Salvatore

A2 - Skavhaug, Amund

PB - Institute of Electrical and Electronics Engineers

Y2 - 1 September 2021 through 3 September 2021

ER -

NMPO: Near-Memory Computing Profiling and Offloading

Abstract

Conference

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

NeMeCo

Best student paper award DSD 2021

Cite this

NMPO: Near-Memory Computing Profiling and Offloading

Abstract

Conference

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Projects

NeMeCo

Prizes

Best student paper award DSD 2021

Cite this