Up by their bootstraps

Online learning in Artificial Neural Networks for CMP uncore power management

Jae Yeon Won, Xi Chen, Paul Gratz, Jiang Hu, Vassos Soteriou Soteriou

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

    Abstract

    With increasing core counts in Chip Multi-Processor (CMP) designs, the size of the on-chip communication fabric and shared Last-Level Caches (LLC), which we term uncore here, is also growing, consuming as much as 30% of die area and a significant portion of chip power budget. In this work, we focus on improving the uncore energy-efficiency using dynamic voltage and frequency scaling. Previous approaches are mostly restricted to reactive techniques, which may respond poorly to abrupt workload and uncore utility changes. We find, however, there are predictable patterns in uncore utility which point towards the potential of a proactive approach to uncore power management. In this work, we utilize artificial intelligence principles to proactively leverage uncore utility pattern prediction via an Artificial Neural Network (ANN). ANNs, however, require training to produce accurate predictions. Architecting an efficient training mechanism without a priori knowledge of the workload is a major challenge. We propose a novel technique in which a simple Proportional Integral (PI) controller is used as a secondary classifier during ANN training, dynamically pulling the ANN up by its bootstraps to achieve accurate predictions. Both the ANN and the PI controller, then, work in tandem once the ANN training phase is complete. The advantage of using a PI controller to initially train the ANN is a dramatic acceleration of the ANN's initial learning phase. Thus, in a real system, this scenario allows quick power-control adaptation to rapid application phase changes and context switches during execution. We show that the proposed technique produces results comparable to those of pure offline training without a need for prerecorded training sets. Full system simulations using the PARSEC benchmark suite show that the bootstrapped ANN improves the energy-delay product of the uncore system by 27% versus existing state-of-the-art methodologies.

    Original languageEnglish (US)
    Title of host publication20th IEEE International Symposium on High Performance Computer Architecture, HPCA 2014
    PublisherIEEE Computer Society
    Pages308-319
    Number of pages12
    ISBN (Print)9781479930975
    DOIs
    StatePublished - Jan 1 2014
    Event20th IEEE International Symposium on High Performance Computer Architecture, HPCA 2014 - Orlando, FL, United States
    Duration: Feb 15 2014Feb 19 2014

    Other

    Other20th IEEE International Symposium on High Performance Computer Architecture, HPCA 2014
    CountryUnited States
    CityOrlando, FL
    Period2/15/142/19/14

    Fingerprint

    Neural networks
    Controllers
    Power management
    Power control
    Artificial intelligence
    Energy efficiency
    Classifiers
    Switches
    Communication

    ASJC Scopus subject areas

    • Hardware and Architecture

    Cite this

    Won, J. Y., Chen, X., Gratz, P., Hu, J., & Soteriou, V. S. (2014). Up by their bootstraps: Online learning in Artificial Neural Networks for CMP uncore power management. In 20th IEEE International Symposium on High Performance Computer Architecture, HPCA 2014 (pp. 308-319). [6835941] IEEE Computer Society. https://doi.org/10.1109/HPCA.2014.6835941

    Up by their bootstraps : Online learning in Artificial Neural Networks for CMP uncore power management. / Won, Jae Yeon; Chen, Xi; Gratz, Paul; Hu, Jiang; Soteriou, Vassos Soteriou.

    20th IEEE International Symposium on High Performance Computer Architecture, HPCA 2014. IEEE Computer Society, 2014. p. 308-319 6835941.

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

    Won, JY, Chen, X, Gratz, P, Hu, J & Soteriou, VS 2014, Up by their bootstraps: Online learning in Artificial Neural Networks for CMP uncore power management. in 20th IEEE International Symposium on High Performance Computer Architecture, HPCA 2014., 6835941, IEEE Computer Society, pp. 308-319, 20th IEEE International Symposium on High Performance Computer Architecture, HPCA 2014, Orlando, FL, United States, 2/15/14. https://doi.org/10.1109/HPCA.2014.6835941
    Won JY, Chen X, Gratz P, Hu J, Soteriou VS. Up by their bootstraps: Online learning in Artificial Neural Networks for CMP uncore power management. In 20th IEEE International Symposium on High Performance Computer Architecture, HPCA 2014. IEEE Computer Society. 2014. p. 308-319. 6835941 https://doi.org/10.1109/HPCA.2014.6835941
    Won, Jae Yeon ; Chen, Xi ; Gratz, Paul ; Hu, Jiang ; Soteriou, Vassos Soteriou. / Up by their bootstraps : Online learning in Artificial Neural Networks for CMP uncore power management. 20th IEEE International Symposium on High Performance Computer Architecture, HPCA 2014. IEEE Computer Society, 2014. pp. 308-319
    @inproceedings{b4e74adc9e794c00ab7b8a35ea8631da,
    title = "Up by their bootstraps: Online learning in Artificial Neural Networks for CMP uncore power management",
    abstract = "With increasing core counts in Chip Multi-Processor (CMP) designs, the size of the on-chip communication fabric and shared Last-Level Caches (LLC), which we term uncore here, is also growing, consuming as much as 30{\%} of die area and a significant portion of chip power budget. In this work, we focus on improving the uncore energy-efficiency using dynamic voltage and frequency scaling. Previous approaches are mostly restricted to reactive techniques, which may respond poorly to abrupt workload and uncore utility changes. We find, however, there are predictable patterns in uncore utility which point towards the potential of a proactive approach to uncore power management. In this work, we utilize artificial intelligence principles to proactively leverage uncore utility pattern prediction via an Artificial Neural Network (ANN). ANNs, however, require training to produce accurate predictions. Architecting an efficient training mechanism without a priori knowledge of the workload is a major challenge. We propose a novel technique in which a simple Proportional Integral (PI) controller is used as a secondary classifier during ANN training, dynamically pulling the ANN up by its bootstraps to achieve accurate predictions. Both the ANN and the PI controller, then, work in tandem once the ANN training phase is complete. The advantage of using a PI controller to initially train the ANN is a dramatic acceleration of the ANN's initial learning phase. Thus, in a real system, this scenario allows quick power-control adaptation to rapid application phase changes and context switches during execution. We show that the proposed technique produces results comparable to those of pure offline training without a need for prerecorded training sets. Full system simulations using the PARSEC benchmark suite show that the bootstrapped ANN improves the energy-delay product of the uncore system by 27{\%} versus existing state-of-the-art methodologies.",
    author = "Won, {Jae Yeon} and Xi Chen and Paul Gratz and Jiang Hu and Soteriou, {Vassos Soteriou}",
    year = "2014",
    month = "1",
    day = "1",
    doi = "10.1109/HPCA.2014.6835941",
    language = "English (US)",
    isbn = "9781479930975",
    pages = "308--319",
    booktitle = "20th IEEE International Symposium on High Performance Computer Architecture, HPCA 2014",
    publisher = "IEEE Computer Society",

    }

    TY - GEN

    T1 - Up by their bootstraps

    T2 - Online learning in Artificial Neural Networks for CMP uncore power management

    AU - Won, Jae Yeon

    AU - Chen, Xi

    AU - Gratz, Paul

    AU - Hu, Jiang

    AU - Soteriou, Vassos Soteriou

    PY - 2014/1/1

    Y1 - 2014/1/1

    N2 - With increasing core counts in Chip Multi-Processor (CMP) designs, the size of the on-chip communication fabric and shared Last-Level Caches (LLC), which we term uncore here, is also growing, consuming as much as 30% of die area and a significant portion of chip power budget. In this work, we focus on improving the uncore energy-efficiency using dynamic voltage and frequency scaling. Previous approaches are mostly restricted to reactive techniques, which may respond poorly to abrupt workload and uncore utility changes. We find, however, there are predictable patterns in uncore utility which point towards the potential of a proactive approach to uncore power management. In this work, we utilize artificial intelligence principles to proactively leverage uncore utility pattern prediction via an Artificial Neural Network (ANN). ANNs, however, require training to produce accurate predictions. Architecting an efficient training mechanism without a priori knowledge of the workload is a major challenge. We propose a novel technique in which a simple Proportional Integral (PI) controller is used as a secondary classifier during ANN training, dynamically pulling the ANN up by its bootstraps to achieve accurate predictions. Both the ANN and the PI controller, then, work in tandem once the ANN training phase is complete. The advantage of using a PI controller to initially train the ANN is a dramatic acceleration of the ANN's initial learning phase. Thus, in a real system, this scenario allows quick power-control adaptation to rapid application phase changes and context switches during execution. We show that the proposed technique produces results comparable to those of pure offline training without a need for prerecorded training sets. Full system simulations using the PARSEC benchmark suite show that the bootstrapped ANN improves the energy-delay product of the uncore system by 27% versus existing state-of-the-art methodologies.

    AB - With increasing core counts in Chip Multi-Processor (CMP) designs, the size of the on-chip communication fabric and shared Last-Level Caches (LLC), which we term uncore here, is also growing, consuming as much as 30% of die area and a significant portion of chip power budget. In this work, we focus on improving the uncore energy-efficiency using dynamic voltage and frequency scaling. Previous approaches are mostly restricted to reactive techniques, which may respond poorly to abrupt workload and uncore utility changes. We find, however, there are predictable patterns in uncore utility which point towards the potential of a proactive approach to uncore power management. In this work, we utilize artificial intelligence principles to proactively leverage uncore utility pattern prediction via an Artificial Neural Network (ANN). ANNs, however, require training to produce accurate predictions. Architecting an efficient training mechanism without a priori knowledge of the workload is a major challenge. We propose a novel technique in which a simple Proportional Integral (PI) controller is used as a secondary classifier during ANN training, dynamically pulling the ANN up by its bootstraps to achieve accurate predictions. Both the ANN and the PI controller, then, work in tandem once the ANN training phase is complete. The advantage of using a PI controller to initially train the ANN is a dramatic acceleration of the ANN's initial learning phase. Thus, in a real system, this scenario allows quick power-control adaptation to rapid application phase changes and context switches during execution. We show that the proposed technique produces results comparable to those of pure offline training without a need for prerecorded training sets. Full system simulations using the PARSEC benchmark suite show that the bootstrapped ANN improves the energy-delay product of the uncore system by 27% versus existing state-of-the-art methodologies.

    UR - http://www.scopus.com/inward/record.url?scp=84903957187&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84903957187&partnerID=8YFLogxK

    U2 - 10.1109/HPCA.2014.6835941

    DO - 10.1109/HPCA.2014.6835941

    M3 - Conference contribution

    SN - 9781479930975

    SP - 308

    EP - 319

    BT - 20th IEEE International Symposium on High Performance Computer Architecture, HPCA 2014

    PB - IEEE Computer Society

    ER -