An Online Learning Methodology for Performance Modeling of Graphics Processors

Ujjwal Gupta, Manoj Babu, Raid Ayoub, Michael Kishinevsky, Francesco Paterna, Suat Gumussoy, Umit Ogras

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Approximately 18% of the 3.2 million smartphone applications rely on integrated graphics processing units (GPUs) to achieve competitive performance. Graphics performance, typically measured in frames per second, is a strong function of the GPU frequency, which in turn has a significant impact on mobile processor power consumption. Consequently, dynamic power management algorithms have to assess the performance sensitivity to the frequency accurately to choose the operating frequency of the GPU effectively. Since the impact of GPU frequency on performance varies rapidly over time, there is a need for online performance models that can adapt to varying workloads. This paper presents a light-weight adaptive runtime performance model that predicts the frame processing time of graphics workloads at runtime without apriori characterization. We employ this model to estimate the frame time sensitivity to the GPU frequency, i.e., the partial derivative of the frame time with respect to the GPU frequency. The proposed model does not rely on any parameter learned offline. Our experiments on the Intel Minnowboard MAX platform running common GPU benchmarks show that the mean absolute percentage error in frame time and frame time sensitivity prediction are 4.2% and 6.7%, respectively.

Original languageEnglish (US)
JournalIEEE Transactions on Computers
DOIs
StateAccepted/In press - May 24 2018

Fingerprint

Graphics Processors
Performance Modeling
Online Learning
Graphics Processing Unit
Methodology
Performance Model
Workload
Power Management
Smartphones
Partial derivative
Graphics processing unit
Power Consumption
Percentage
Electric power utilization
Choose
Vary
Benchmark
Derivatives
Predict
Prediction

Keywords

  • Adaptation models
  • Computational modeling
  • frequency scaling
  • Graphics processing units
  • Heuristic algorithms
  • Integrated GPUs
  • online learning
  • performance modeling
  • power management
  • Power system management
  • Predictive models
  • RLS
  • Time-frequency analysis

ASJC Scopus subject areas

  • Software
  • Theoretical Computer Science
  • Hardware and Architecture
  • Computational Theory and Mathematics

Cite this

An Online Learning Methodology for Performance Modeling of Graphics Processors. / Gupta, Ujjwal; Babu, Manoj; Ayoub, Raid; Kishinevsky, Michael; Paterna, Francesco; Gumussoy, Suat; Ogras, Umit.

In: IEEE Transactions on Computers, 24.05.2018.

Research output: Contribution to journalArticle

Gupta, Ujjwal ; Babu, Manoj ; Ayoub, Raid ; Kishinevsky, Michael ; Paterna, Francesco ; Gumussoy, Suat ; Ogras, Umit. / An Online Learning Methodology for Performance Modeling of Graphics Processors. In: IEEE Transactions on Computers. 2018.
@article{dca1c7c027fd4eb7908e2f9d078f8741,
title = "An Online Learning Methodology for Performance Modeling of Graphics Processors",
abstract = "Approximately 18{\%} of the 3.2 million smartphone applications rely on integrated graphics processing units (GPUs) to achieve competitive performance. Graphics performance, typically measured in frames per second, is a strong function of the GPU frequency, which in turn has a significant impact on mobile processor power consumption. Consequently, dynamic power management algorithms have to assess the performance sensitivity to the frequency accurately to choose the operating frequency of the GPU effectively. Since the impact of GPU frequency on performance varies rapidly over time, there is a need for online performance models that can adapt to varying workloads. This paper presents a light-weight adaptive runtime performance model that predicts the frame processing time of graphics workloads at runtime without apriori characterization. We employ this model to estimate the frame time sensitivity to the GPU frequency, i.e., the partial derivative of the frame time with respect to the GPU frequency. The proposed model does not rely on any parameter learned offline. Our experiments on the Intel Minnowboard MAX platform running common GPU benchmarks show that the mean absolute percentage error in frame time and frame time sensitivity prediction are 4.2{\%} and 6.7{\%}, respectively.",
keywords = "Adaptation models, Computational modeling, frequency scaling, Graphics processing units, Heuristic algorithms, Integrated GPUs, online learning, performance modeling, power management, Power system management, Predictive models, RLS, Time-frequency analysis",
author = "Ujjwal Gupta and Manoj Babu and Raid Ayoub and Michael Kishinevsky and Francesco Paterna and Suat Gumussoy and Umit Ogras",
year = "2018",
month = "5",
day = "24",
doi = "10.1109/TC.2018.2840710",
language = "English (US)",
journal = "IEEE Transactions on Computers",
issn = "0018-9340",
publisher = "IEEE Computer Society",

}

TY - JOUR

T1 - An Online Learning Methodology for Performance Modeling of Graphics Processors

AU - Gupta, Ujjwal

AU - Babu, Manoj

AU - Ayoub, Raid

AU - Kishinevsky, Michael

AU - Paterna, Francesco

AU - Gumussoy, Suat

AU - Ogras, Umit

PY - 2018/5/24

Y1 - 2018/5/24

N2 - Approximately 18% of the 3.2 million smartphone applications rely on integrated graphics processing units (GPUs) to achieve competitive performance. Graphics performance, typically measured in frames per second, is a strong function of the GPU frequency, which in turn has a significant impact on mobile processor power consumption. Consequently, dynamic power management algorithms have to assess the performance sensitivity to the frequency accurately to choose the operating frequency of the GPU effectively. Since the impact of GPU frequency on performance varies rapidly over time, there is a need for online performance models that can adapt to varying workloads. This paper presents a light-weight adaptive runtime performance model that predicts the frame processing time of graphics workloads at runtime without apriori characterization. We employ this model to estimate the frame time sensitivity to the GPU frequency, i.e., the partial derivative of the frame time with respect to the GPU frequency. The proposed model does not rely on any parameter learned offline. Our experiments on the Intel Minnowboard MAX platform running common GPU benchmarks show that the mean absolute percentage error in frame time and frame time sensitivity prediction are 4.2% and 6.7%, respectively.

AB - Approximately 18% of the 3.2 million smartphone applications rely on integrated graphics processing units (GPUs) to achieve competitive performance. Graphics performance, typically measured in frames per second, is a strong function of the GPU frequency, which in turn has a significant impact on mobile processor power consumption. Consequently, dynamic power management algorithms have to assess the performance sensitivity to the frequency accurately to choose the operating frequency of the GPU effectively. Since the impact of GPU frequency on performance varies rapidly over time, there is a need for online performance models that can adapt to varying workloads. This paper presents a light-weight adaptive runtime performance model that predicts the frame processing time of graphics workloads at runtime without apriori characterization. We employ this model to estimate the frame time sensitivity to the GPU frequency, i.e., the partial derivative of the frame time with respect to the GPU frequency. The proposed model does not rely on any parameter learned offline. Our experiments on the Intel Minnowboard MAX platform running common GPU benchmarks show that the mean absolute percentage error in frame time and frame time sensitivity prediction are 4.2% and 6.7%, respectively.

KW - Adaptation models

KW - Computational modeling

KW - frequency scaling

KW - Graphics processing units

KW - Heuristic algorithms

KW - Integrated GPUs

KW - online learning

KW - performance modeling

KW - power management

KW - Power system management

KW - Predictive models

KW - RLS

KW - Time-frequency analysis

UR - http://www.scopus.com/inward/record.url?scp=85047613199&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85047613199&partnerID=8YFLogxK

U2 - 10.1109/TC.2018.2840710

DO - 10.1109/TC.2018.2840710

M3 - Article

AN - SCOPUS:85047613199

JO - IEEE Transactions on Computers

JF - IEEE Transactions on Computers

SN - 0018-9340

ER -