Soft-NEON: A study on replacing the NEON engine of an ARM SoC with a reconfigurable fabric

Jose Raul Garcia Ordaz; Dirk Koch

doi:10.1109/ASAP.2016.7760802

Soft-NEON: A study on replacing the NEON engine of an ARM SoC with a reconfigurable fabric

Jose Raul Garcia Ordaz
, Dirk Koch

Advanced Processor Technology

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

459 Downloads (Pure)

Abstract

Power is a limiting factor in the design of embedded processors. For this reason adding more instruction extensions is not a scalable option. To overcome this issue, we study the effects of replacing the NEON unit of an ARM SoC with an FPGA-like reconfigurable fabric. We measure the gap between the conventional hard-NEON and a soft-NEON implementation. We found that the soft-NEON has an overhead of 25.17× and 6.23× for area and latency, respectively. This overhead is reduced by exploiting the reconfigurability of the fabric by incorporating FPGA-specific optimization techniques. Moreover, we show that instead of implementing the pre-defined NEON instruction set, custom instructions can be loaded to the reconfigurable fabric by using a HLS compilation flow. With this approach performance gains of over 2.8× have been obtained for some kernels.

Original language	English
Title of host publication	2016 IEEE 27th International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2016
Publisher	IEEE
Pages	229-230
Number of pages	2
Volume	2016-November
ISBN (Electronic)	9781509015030
DOIs	https://doi.org/10.1109/ASAP.2016.7760802
Publication status	Published - 28 Nov 2016
Event	27th IEEE International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2016 - London, United Kingdom Duration: 6 Jul 2016 → 8 Jul 2016

Conference

Conference	27th IEEE International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2016
Country/Territory	United Kingdom
City	London
Period	6/07/16 → 8/07/16

Access to Document

10.1109/ASAP.2016.7760802

ASAP2016_soft_NeonAccepted author manuscript, 88.7 KB

Cite this

@inproceedings{6dc297ed32f64b4dbacbff04296097d3,

title = "Soft-NEON: A study on replacing the NEON engine of an ARM SoC with a reconfigurable fabric",

abstract = "Power is a limiting factor in the design of embedded processors. For this reason adding more instruction extensions is not a scalable option. To overcome this issue, we study the effects of replacing the NEON unit of an ARM SoC with an FPGA-like reconfigurable fabric. We measure the gap between the conventional hard-NEON and a soft-NEON implementation. We found that the soft-NEON has an overhead of 25.17{\texttimes} and 6.23{\texttimes} for area and latency, respectively. This overhead is reduced by exploiting the reconfigurability of the fabric by incorporating FPGA-specific optimization techniques. Moreover, we show that instead of implementing the pre-defined NEON instruction set, custom instructions can be loaded to the reconfigurable fabric by using a HLS compilation flow. With this approach performance gains of over 2.8{\texttimes} have been obtained for some kernels.",

author = "\{Garcia Ordaz\}, \{Jose Raul\} and Dirk Koch",

year = "2016",

month = nov,

day = "28",

doi = "10.1109/ASAP.2016.7760802",

language = "English",

volume = "2016-November",

pages = "229--230",

booktitle = "2016 IEEE 27th International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2016",

publisher = "IEEE",

address = "United States",

note = "27th IEEE International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2016 ; Conference date: 06-07-2016 Through 08-07-2016",

}

Garcia Ordaz, JR & Koch, D 2016, Soft-NEON: A study on replacing the NEON engine of an ARM SoC with a reconfigurable fabric. in 2016 IEEE 27th International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2016. vol. 2016-November, 7760802, IEEE, pp. 229-230, 27th IEEE International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2016, London, United Kingdom, 6/07/16. https://doi.org/10.1109/ASAP.2016.7760802

Soft-NEON: A study on replacing the NEON engine of an ARM SoC with a reconfigurable fabric. / Garcia Ordaz, Jose Raul; Koch, Dirk.
2016 IEEE 27th International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2016. Vol. 2016-November IEEE, 2016. p. 229-230 7760802.

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

TY - GEN

T1 - Soft-NEON

T2 - 27th IEEE International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2016

AU - Garcia Ordaz, Jose Raul

AU - Koch, Dirk

PY - 2016/11/28

Y1 - 2016/11/28

N2 - Power is a limiting factor in the design of embedded processors. For this reason adding more instruction extensions is not a scalable option. To overcome this issue, we study the effects of replacing the NEON unit of an ARM SoC with an FPGA-like reconfigurable fabric. We measure the gap between the conventional hard-NEON and a soft-NEON implementation. We found that the soft-NEON has an overhead of 25.17× and 6.23× for area and latency, respectively. This overhead is reduced by exploiting the reconfigurability of the fabric by incorporating FPGA-specific optimization techniques. Moreover, we show that instead of implementing the pre-defined NEON instruction set, custom instructions can be loaded to the reconfigurable fabric by using a HLS compilation flow. With this approach performance gains of over 2.8× have been obtained for some kernels.

AB - Power is a limiting factor in the design of embedded processors. For this reason adding more instruction extensions is not a scalable option. To overcome this issue, we study the effects of replacing the NEON unit of an ARM SoC with an FPGA-like reconfigurable fabric. We measure the gap between the conventional hard-NEON and a soft-NEON implementation. We found that the soft-NEON has an overhead of 25.17× and 6.23× for area and latency, respectively. This overhead is reduced by exploiting the reconfigurability of the fabric by incorporating FPGA-specific optimization techniques. Moreover, we show that instead of implementing the pre-defined NEON instruction set, custom instructions can be loaded to the reconfigurable fabric by using a HLS compilation flow. With this approach performance gains of over 2.8× have been obtained for some kernels.

UR - https://www.scopus.com/pages/publications/85006852919

U2 - 10.1109/ASAP.2016.7760802

DO - 10.1109/ASAP.2016.7760802

M3 - Conference contribution

AN - SCOPUS:85006852919

VL - 2016-November

SP - 229

EP - 230

BT - 2016 IEEE 27th International Conference on Application-Specific Systems, Architectures and Processors, ASAP 2016

PB - IEEE

Y2 - 6 July 2016 through 8 July 2016

ER -

Soft-NEON: A study on replacing the NEON engine of an ARM SoC with a reconfigurable fabric

Abstract

Conference

Access to Document

Other files and links

Fingerprint

Cite this