Hybrid compile and run-time memory management for a 3D-stacked reconfigurable accelerator

Lovic Gauthier, Shinya Ueno, Inoue Koji

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

Abstract

This paper presents a hybrid compile and run-time memory management technique for a 3D-stacked reconfigurable accelerator including a memory layer composed of multiple memory units whose parallel access allows a very high bandwidth. The technique inserts allocation, free and data transfers into the code for using the memory layer and avoids memory overflows by adding a limited number of additional copies to and from the host memory. When compile-time information is lacking, the technique relies on run-time decisions for controlling these memory operations. Experiments show that, compared to a pessimistic approach, the overhead for avoiding overflows can be cut on average by 27%, 45% and 63% when the size of each memory unit is respectively 1kB, 128kB and 1MB.

Original languageEnglish
Title of host publication2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013
PublisherIEEE Computer Society
ISBN (Print)9781479914005
DOIs
Publication statusPublished - Jan 1 2013
Event2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013 - Montreal, QC, Canada
Duration: Sep 29 2013Oct 4 2013

Publication series

Name2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013

Other

Other2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013
CountryCanada
CityMontreal, QC
Period9/29/1310/4/13

Fingerprint

Particle accelerators
Data storage equipment
Data transfer
Bandwidth
Experiments

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture
  • Control and Systems Engineering

Cite this

Gauthier, L., Ueno, S., & Koji, I. (2013). Hybrid compile and run-time memory management for a 3D-stacked reconfigurable accelerator. In 2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013 [6662514] (2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013). IEEE Computer Society. https://doi.org/10.1109/CASES.2013.6662514

Hybrid compile and run-time memory management for a 3D-stacked reconfigurable accelerator. / Gauthier, Lovic; Ueno, Shinya; Koji, Inoue.

2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013. IEEE Computer Society, 2013. 6662514 (2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013).

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

Gauthier, L, Ueno, S & Koji, I 2013, Hybrid compile and run-time memory management for a 3D-stacked reconfigurable accelerator. in 2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013., 6662514, 2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013, IEEE Computer Society, 2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013, Montreal, QC, Canada, 9/29/13. https://doi.org/10.1109/CASES.2013.6662514
Gauthier L, Ueno S, Koji I. Hybrid compile and run-time memory management for a 3D-stacked reconfigurable accelerator. In 2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013. IEEE Computer Society. 2013. 6662514. (2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013). https://doi.org/10.1109/CASES.2013.6662514
Gauthier, Lovic ; Ueno, Shinya ; Koji, Inoue. / Hybrid compile and run-time memory management for a 3D-stacked reconfigurable accelerator. 2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013. IEEE Computer Society, 2013. (2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013).
@inproceedings{0c8c8681eef04481b3555df7d33d76ef,
title = "Hybrid compile and run-time memory management for a 3D-stacked reconfigurable accelerator",
abstract = "This paper presents a hybrid compile and run-time memory management technique for a 3D-stacked reconfigurable accelerator including a memory layer composed of multiple memory units whose parallel access allows a very high bandwidth. The technique inserts allocation, free and data transfers into the code for using the memory layer and avoids memory overflows by adding a limited number of additional copies to and from the host memory. When compile-time information is lacking, the technique relies on run-time decisions for controlling these memory operations. Experiments show that, compared to a pessimistic approach, the overhead for avoiding overflows can be cut on average by 27{\%}, 45{\%} and 63{\%} when the size of each memory unit is respectively 1kB, 128kB and 1MB.",
author = "Lovic Gauthier and Shinya Ueno and Inoue Koji",
year = "2013",
month = "1",
day = "1",
doi = "10.1109/CASES.2013.6662514",
language = "English",
isbn = "9781479914005",
series = "2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013",
publisher = "IEEE Computer Society",
booktitle = "2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013",
address = "United States",

}

TY - GEN

T1 - Hybrid compile and run-time memory management for a 3D-stacked reconfigurable accelerator

AU - Gauthier, Lovic

AU - Ueno, Shinya

AU - Koji, Inoue

PY - 2013/1/1

Y1 - 2013/1/1

N2 - This paper presents a hybrid compile and run-time memory management technique for a 3D-stacked reconfigurable accelerator including a memory layer composed of multiple memory units whose parallel access allows a very high bandwidth. The technique inserts allocation, free and data transfers into the code for using the memory layer and avoids memory overflows by adding a limited number of additional copies to and from the host memory. When compile-time information is lacking, the technique relies on run-time decisions for controlling these memory operations. Experiments show that, compared to a pessimistic approach, the overhead for avoiding overflows can be cut on average by 27%, 45% and 63% when the size of each memory unit is respectively 1kB, 128kB and 1MB.

AB - This paper presents a hybrid compile and run-time memory management technique for a 3D-stacked reconfigurable accelerator including a memory layer composed of multiple memory units whose parallel access allows a very high bandwidth. The technique inserts allocation, free and data transfers into the code for using the memory layer and avoids memory overflows by adding a limited number of additional copies to and from the host memory. When compile-time information is lacking, the technique relies on run-time decisions for controlling these memory operations. Experiments show that, compared to a pessimistic approach, the overhead for avoiding overflows can be cut on average by 27%, 45% and 63% when the size of each memory unit is respectively 1kB, 128kB and 1MB.

UR - http://www.scopus.com/inward/record.url?scp=84892638573&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84892638573&partnerID=8YFLogxK

U2 - 10.1109/CASES.2013.6662514

DO - 10.1109/CASES.2013.6662514

M3 - Conference contribution

AN - SCOPUS:84892638573

SN - 9781479914005

T3 - 2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013

BT - 2013 International Conference on Compilers, Architecture and Synthesis for Embedded Systems, CASES 2013

PB - IEEE Computer Society

ER -