Approximating HW Accelerators through Partial Extractions onto Shared Artificial Neural Networks

Prattay Chowdhury; Jorge Castro Godínez; Benjamin Carrion Schafer

doi:10.1145/3566097.3567887

Approximating HW Accelerators through Partial Extractions onto Shared Artificial Neural Networks

Prattay Chowdhury, Jorge Castro Godínez, Benjamin Carrion Schafer

Escuela de Ingeniería Electrónica

University of Texas at Dallas

Producción científica: Capítulo del libro/informe/acta de congreso › Contribución a la conferencia › revisión exhaustiva

Resumen

One approach that has been suggested to further reduce the energy consumption of heterogenous Systems-on-Chip (SoCs) is approximate computing. In approximate computing the error at the output is relaxed in order to simplify the hardware and thus, achieve lowerpower. Fortunately, most of the hardware accelerators in these SoCs are also amenable to approximate computing.
In this work we propose a fully automatic method that substitutes portions of a hardware accelerator specified in C/C++/SystemC for
High-Level Synthesis (HLS) to an Artificial Neural Network (ANN).
ANNs have many advantages that make them well suited for this.
First, they are very scalable which allows to approximate multiple separate portions of the behavioral description simultaneously on them. Second, multiple ANNs can be fused together and reoptimized to further reduce the power consumption. We use this to share the ANN to approximate multiple different HW accelerators in the same SoC . Experimental results with different error thresholds show that our proposed approach leads to better results than the state of the art.

Idioma original	Inglés
Título de la publicación alojada	ASPDAC '23: Proceedings of the 28th Asia and South Pacific Design Automation Conference
Lugar de publicación	Tokyo, Japan
Editorial	Institute of Electrical and Electronics Engineers Inc.
Páginas	410-415
Número de páginas	6
ISBN (versión digital)	9781450397834
DOI	https://doi.org/10.1145/3566097.3567887
Estado	Publicada - 16 ene 2023
Evento	28th Asia and South Pacific Design Automation Conference, ASP-DAC 2023 - Tokyo, Japón Duración: 16 ene 2023 → 19 ene 2023

Serie de la publicación

Nombre	Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

Conferencia

Conferencia	28th Asia and South Pacific Design Automation Conference, ASP-DAC 2023
País/Territorio	Japón
Ciudad	Tokyo
Período	16/01/23 → 19/01/23

ODS de las Naciones Unidas

Este resultado contribuye a los siguientes Objetivos de Desarrollo Sostenible

Acceder al documento

10.1145/3566097.3567887

Otros archivos y enlaces

Enlace a la publicación en Scopus

Citar esto

Chowdhury, P., Godínez, J. C., & Schafer, B. C. (2023). Approximating HW Accelerators through Partial Extractions onto Shared Artificial Neural Networks. En ASPDAC '23: Proceedings of the 28th Asia and South Pacific Design Automation Conference (pp. 410-415). (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1145/3566097.3567887

Chowdhury, Prattay ; Godínez, Jorge Castro ; Schafer, Benjamin Carrion. / Approximating HW Accelerators through Partial Extractions onto Shared Artificial Neural Networks. ASPDAC '23: Proceedings of the 28th Asia and South Pacific Design Automation Conference. Tokyo, Japan : Institute of Electrical and Electronics Engineers Inc., 2023. pp. 410-415 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC).

@inproceedings{2e066714fe884b67976b615e6b5068b8,

title = "Approximating HW Accelerators through Partial Extractions onto Shared Artificial Neural Networks",

abstract = "One approach that has been suggested to further reduce the energy consumption of heterogenous Systems-on-Chip (SoCs) is approximate computing. In approximate computing the error at the output is relaxed in order to simplify the hardware and thus, achieve lowerpower. Fortunately, most of the hardware accelerators in these SoCs are also amenable to approximate computing.In this work we propose a fully automatic method that substitutes portions of a hardware accelerator specified in C/C++/SystemC forHigh-Level Synthesis (HLS) to an Artificial Neural Network (ANN).ANNs have many advantages that make them well suited for this.First, they are very scalable which allows to approximate multiple separate portions of the behavioral description simultaneously on them. Second, multiple ANNs can be fused together and reoptimized to further reduce the power consumption. We use this to share the ANN to approximate multiple different HW accelerators in the same SoC . Experimental results with different error thresholds show that our proposed approach leads to better results than the state of the art.",

author = "Prattay Chowdhury and God{\'i}nez, {Jorge Castro} and Schafer, {Benjamin Carrion}",

note = "Publisher Copyright: {\textcopyright} 2023 Copyright held by the owner/author(s).; 28th Asia and South Pacific Design Automation Conference, ASP-DAC 2023 ; Conference date: 16-01-2023 Through 19-01-2023",

year = "2023",

month = jan,

day = "16",

doi = "10.1145/3566097.3567887",

language = "Ingl{\'e}s",

series = "Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "410--415",

booktitle = "ASPDAC '23: Proceedings of the 28th Asia and South Pacific Design Automation Conference",

}

Chowdhury, P, Godínez, JC & Schafer, BC 2023, Approximating HW Accelerators through Partial Extractions onto Shared Artificial Neural Networks. En ASPDAC '23: Proceedings of the 28th Asia and South Pacific Design Automation Conference. Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC, Institute of Electrical and Electronics Engineers Inc., Tokyo, Japan, pp. 410-415, 28th Asia and South Pacific Design Automation Conference, ASP-DAC 2023, Tokyo, Japón, 16/01/23. https://doi.org/10.1145/3566097.3567887

Approximating HW Accelerators through Partial Extractions onto Shared Artificial Neural Networks. / Chowdhury, Prattay; Godínez, Jorge Castro; Schafer, Benjamin Carrion.
ASPDAC '23: Proceedings of the 28th Asia and South Pacific Design Automation Conference. Tokyo, Japan: Institute of Electrical and Electronics Engineers Inc., 2023. p. 410-415 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC).

Producción científica: Capítulo del libro/informe/acta de congreso › Contribución a la conferencia › revisión exhaustiva

TY - GEN

T1 - Approximating HW Accelerators through Partial Extractions onto Shared Artificial Neural Networks

AU - Chowdhury, Prattay

AU - Godínez, Jorge Castro

AU - Schafer, Benjamin Carrion

PY - 2023/1/16

Y1 - 2023/1/16

N2 - One approach that has been suggested to further reduce the energy consumption of heterogenous Systems-on-Chip (SoCs) is approximate computing. In approximate computing the error at the output is relaxed in order to simplify the hardware and thus, achieve lowerpower. Fortunately, most of the hardware accelerators in these SoCs are also amenable to approximate computing.In this work we propose a fully automatic method that substitutes portions of a hardware accelerator specified in C/C++/SystemC forHigh-Level Synthesis (HLS) to an Artificial Neural Network (ANN).ANNs have many advantages that make them well suited for this.First, they are very scalable which allows to approximate multiple separate portions of the behavioral description simultaneously on them. Second, multiple ANNs can be fused together and reoptimized to further reduce the power consumption. We use this to share the ANN to approximate multiple different HW accelerators in the same SoC . Experimental results with different error thresholds show that our proposed approach leads to better results than the state of the art.

AB - One approach that has been suggested to further reduce the energy consumption of heterogenous Systems-on-Chip (SoCs) is approximate computing. In approximate computing the error at the output is relaxed in order to simplify the hardware and thus, achieve lowerpower. Fortunately, most of the hardware accelerators in these SoCs are also amenable to approximate computing.In this work we propose a fully automatic method that substitutes portions of a hardware accelerator specified in C/C++/SystemC forHigh-Level Synthesis (HLS) to an Artificial Neural Network (ANN).ANNs have many advantages that make them well suited for this.First, they are very scalable which allows to approximate multiple separate portions of the behavioral description simultaneously on them. Second, multiple ANNs can be fused together and reoptimized to further reduce the power consumption. We use this to share the ANN to approximate multiple different HW accelerators in the same SoC . Experimental results with different error thresholds show that our proposed approach leads to better results than the state of the art.

UR - http://www.scopus.com/inward/record.url?scp=85148486725&partnerID=8YFLogxK

U2 - 10.1145/3566097.3567887

DO - 10.1145/3566097.3567887

M3 - Contribución a la conferencia

AN - SCOPUS:85148486725

T3 - Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

SP - 410

EP - 415

BT - ASPDAC '23: Proceedings of the 28th Asia and South Pacific Design Automation Conference

PB - Institute of Electrical and Electronics Engineers Inc.

CY - Tokyo, Japan

T2 - 28th Asia and South Pacific Design Automation Conference, ASP-DAC 2023

Y2 - 16 January 2023 through 19 January 2023

ER -

Chowdhury P, Godínez JC, Schafer BC. Approximating HW Accelerators through Partial Extractions onto Shared Artificial Neural Networks. En ASPDAC '23: Proceedings of the 28th Asia and South Pacific Design Automation Conference. Tokyo, Japan: Institute of Electrical and Electronics Engineers Inc. 2023. p. 410-415. (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). doi: 10.1145/3566097.3567887

Approximating HW Accelerators through Partial Extractions onto Shared Artificial Neural Networks

Resumen

Serie de la publicación

Conferencia

ODS de las Naciones Unidas

Acceder al documento

Otros archivos y enlaces

Huella

Citar esto