Energy-Efficient Approximate Least Squares Accelerator: A Case Study of Radio Astronomy Calibration Processing

Syed Ghayoor Abbas Gillani, Alexander Krapukhin, Andre B.J. Kokkeler

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

4 Downloads (Pure)

Abstract

Approximate computing allows the introduction of inaccuracy in the computation for cost savings, such as energy consumption, chip-area, and latency. Targeting energy efficiency, approximate designs for multipliers, adders, and multiply-accumulate (MAC) have been extensively investigated in the past decade. However, accelerator designs for relatively bigger architectures have been of less attention yet. The Least Squares (LS) algorithm is widely used in digital signal processing applications, e.g., image reconstruction. This work proposes a novel LS accelerator design based on a heterogeneous architecture, where the heterogeneity is introduced using accurate and approximate processing cores. We have considered a case study of radio astronomy calibration processing that employs a complex-input iterative LS algorithm. Our proposed methodology exploits the intrinsic error-resilience of the aforesaid algorithm, where initial iterations are processed on approximate modules while the later ones on accurate modules. Our energy-quality experiments have shown up to 24% of energy savings as compared to an accurate (optimized) counterpart for biased designs and up to 29% energy savings when unbiasing is introduced. The proposed LS accelerator design does not increase the number of iterations and provides sufficient precision to converge to an acceptable solution.
Original languageEnglish
Title of host publication16th ACM International Conference on Computing Frontiers
PublisherAssociation for Computing Machinery (ACM)
Pages358-365
ISBN (Print)978-1-4503-6685-4
DOIs
Publication statusPublished - 19 May 2019
Event16th ACM International Conference on Computing Frontiers 2019 - Alghero, Italy
Duration: 30 Apr 20192 May 2019
Conference number: 16
http://www.computingfrontiers.org/2019/

Conference

Conference16th ACM International Conference on Computing Frontiers 2019
Abbreviated titleCF 2019
CountryItaly
CityAlghero
Period30/04/192/05/19
Internet address

Fingerprint

Radio astronomy
Particle accelerators
Calibration
Processing
Energy conservation
Adders
Digital signal processing
Image reconstruction
Energy efficiency
Energy utilization
Costs
Experiments

Keywords

  • Least squares accelerator
  • iterative workloads
  • Approximate Computing
  • Energy Efficiency
  • Radio astronomy

Cite this

Gillani, S. G. A., Krapukhin, A., & Kokkeler, A. B. J. (2019). Energy-Efficient Approximate Least Squares Accelerator: A Case Study of Radio Astronomy Calibration Processing. In 16th ACM International Conference on Computing Frontiers (pp. 358-365). Association for Computing Machinery (ACM). https://doi.org/10.1145/3310273.3323161
Gillani, Syed Ghayoor Abbas ; Krapukhin, Alexander ; Kokkeler, Andre B.J. / Energy-Efficient Approximate Least Squares Accelerator : A Case Study of Radio Astronomy Calibration Processing. 16th ACM International Conference on Computing Frontiers. Association for Computing Machinery (ACM), 2019. pp. 358-365
@inproceedings{76b918fc74bb45c8ad31006f1a281034,
title = "Energy-Efficient Approximate Least Squares Accelerator: A Case Study of Radio Astronomy Calibration Processing",
abstract = "Approximate computing allows the introduction of inaccuracy in the computation for cost savings, such as energy consumption, chip-area, and latency. Targeting energy efficiency, approximate designs for multipliers, adders, and multiply-accumulate (MAC) have been extensively investigated in the past decade. However, accelerator designs for relatively bigger architectures have been of less attention yet. The Least Squares (LS) algorithm is widely used in digital signal processing applications, e.g., image reconstruction. This work proposes a novel LS accelerator design based on a heterogeneous architecture, where the heterogeneity is introduced using accurate and approximate processing cores. We have considered a case study of radio astronomy calibration processing that employs a complex-input iterative LS algorithm. Our proposed methodology exploits the intrinsic error-resilience of the aforesaid algorithm, where initial iterations are processed on approximate modules while the later ones on accurate modules. Our energy-quality experiments have shown up to 24{\%} of energy savings as compared to an accurate (optimized) counterpart for biased designs and up to 29{\%} energy savings when unbiasing is introduced. The proposed LS accelerator design does not increase the number of iterations and provides sufficient precision to converge to an acceptable solution.",
keywords = "Least squares accelerator, iterative workloads, Approximate Computing, Energy Efficiency, Radio astronomy",
author = "Gillani, {Syed Ghayoor Abbas} and Alexander Krapukhin and Kokkeler, {Andre B.J.}",
year = "2019",
month = "5",
day = "19",
doi = "10.1145/3310273.3323161",
language = "English",
isbn = "978-1-4503-6685-4",
pages = "358--365",
booktitle = "16th ACM International Conference on Computing Frontiers",
publisher = "Association for Computing Machinery (ACM)",
address = "United States",

}

Gillani, SGA, Krapukhin, A & Kokkeler, ABJ 2019, Energy-Efficient Approximate Least Squares Accelerator: A Case Study of Radio Astronomy Calibration Processing. in 16th ACM International Conference on Computing Frontiers. Association for Computing Machinery (ACM), pp. 358-365, 16th ACM International Conference on Computing Frontiers 2019, Alghero, Italy, 30/04/19. https://doi.org/10.1145/3310273.3323161

Energy-Efficient Approximate Least Squares Accelerator : A Case Study of Radio Astronomy Calibration Processing. / Gillani, Syed Ghayoor Abbas; Krapukhin, Alexander; Kokkeler, Andre B.J.

16th ACM International Conference on Computing Frontiers. Association for Computing Machinery (ACM), 2019. p. 358-365.

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Energy-Efficient Approximate Least Squares Accelerator

T2 - A Case Study of Radio Astronomy Calibration Processing

AU - Gillani, Syed Ghayoor Abbas

AU - Krapukhin, Alexander

AU - Kokkeler, Andre B.J.

PY - 2019/5/19

Y1 - 2019/5/19

N2 - Approximate computing allows the introduction of inaccuracy in the computation for cost savings, such as energy consumption, chip-area, and latency. Targeting energy efficiency, approximate designs for multipliers, adders, and multiply-accumulate (MAC) have been extensively investigated in the past decade. However, accelerator designs for relatively bigger architectures have been of less attention yet. The Least Squares (LS) algorithm is widely used in digital signal processing applications, e.g., image reconstruction. This work proposes a novel LS accelerator design based on a heterogeneous architecture, where the heterogeneity is introduced using accurate and approximate processing cores. We have considered a case study of radio astronomy calibration processing that employs a complex-input iterative LS algorithm. Our proposed methodology exploits the intrinsic error-resilience of the aforesaid algorithm, where initial iterations are processed on approximate modules while the later ones on accurate modules. Our energy-quality experiments have shown up to 24% of energy savings as compared to an accurate (optimized) counterpart for biased designs and up to 29% energy savings when unbiasing is introduced. The proposed LS accelerator design does not increase the number of iterations and provides sufficient precision to converge to an acceptable solution.

AB - Approximate computing allows the introduction of inaccuracy in the computation for cost savings, such as energy consumption, chip-area, and latency. Targeting energy efficiency, approximate designs for multipliers, adders, and multiply-accumulate (MAC) have been extensively investigated in the past decade. However, accelerator designs for relatively bigger architectures have been of less attention yet. The Least Squares (LS) algorithm is widely used in digital signal processing applications, e.g., image reconstruction. This work proposes a novel LS accelerator design based on a heterogeneous architecture, where the heterogeneity is introduced using accurate and approximate processing cores. We have considered a case study of radio astronomy calibration processing that employs a complex-input iterative LS algorithm. Our proposed methodology exploits the intrinsic error-resilience of the aforesaid algorithm, where initial iterations are processed on approximate modules while the later ones on accurate modules. Our energy-quality experiments have shown up to 24% of energy savings as compared to an accurate (optimized) counterpart for biased designs and up to 29% energy savings when unbiasing is introduced. The proposed LS accelerator design does not increase the number of iterations and provides sufficient precision to converge to an acceptable solution.

KW - Least squares accelerator

KW - iterative workloads

KW - Approximate Computing

KW - Energy Efficiency

KW - Radio astronomy

U2 - 10.1145/3310273.3323161

DO - 10.1145/3310273.3323161

M3 - Conference contribution

SN - 978-1-4503-6685-4

SP - 358

EP - 365

BT - 16th ACM International Conference on Computing Frontiers

PB - Association for Computing Machinery (ACM)

ER -

Gillani SGA, Krapukhin A, Kokkeler ABJ. Energy-Efficient Approximate Least Squares Accelerator: A Case Study of Radio Astronomy Calibration Processing. In 16th ACM International Conference on Computing Frontiers. Association for Computing Machinery (ACM). 2019. p. 358-365 https://doi.org/10.1145/3310273.3323161