Real-time multiprocessor architecture for sharing stream processing accelerators

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

    1 Citation (Scopus)
    53 Downloads (Pure)

    Abstract

    Stream processing accelerators are often applied in MPSoCs for software defined radios. Sharing of these accelerators between different streams could improve their utilization and reduce thereby the hardware cost but is challenging under real-time constraints. In this paper we introduce entry- and exit-gateways that are responsible for multiplexing blocks of data over accelerators under real-time constraints. These gateways check for the availability of sufficient data and space and thereby enable the derivation of a dataflow model of the application. The dataflow model is used to verify the worst-case temporal behavior based on the sizes of the blocks of data used for multiplexing. We demonstrate that required buffer capacities are non-monotone in the block size. Therefore, an ILP is presented to compute minimum block sizes and sufficient buffer capacities. The benefits of sharing accelerators are demonstrated using a multi-core system that is implemented on a Virtex 6 FPGA. A stereo audio stream from a PAL video signal is demodulated in this system in real-time where two accelerators are shared within and between two streams. In this system sharing reduces the number of accelerators by 75% and reduced the number of logic cells with 63%.
    Original languageUndefined
    Title of host publication22nd Reconfigurable Architectures Workshop (RAW 2015)
    Place of PublicationUSA
    PublisherIEEE Computer Society
    Pages81-89
    Number of pages9
    ISBN (Print)978-1-4673-7684-6
    DOIs
    Publication statusPublished - 25 May 2015

    Publication series

    Name
    PublisherIEEE Computer Society

    Keywords

    • Real-time applications
    • multiprocessing systems
    • stream processing accelerators
    • METIS-312627
    • IR-96052
    • Data flow
    • accelerator sharing
    • EWI-26067

    Cite this

    Dekens, B. H. J., Bekooij, M. J. G., & Smit, G. J. M. (2015). Real-time multiprocessor architecture for sharing stream processing accelerators. In 22nd Reconfigurable Architectures Workshop (RAW 2015) (pp. 81-89). USA: IEEE Computer Society. https://doi.org/10.1109/IPDPSW.2015.147
    Dekens, B.H.J. ; Bekooij, Marco Jan Gerrit ; Smit, Gerardus Johannes Maria. / Real-time multiprocessor architecture for sharing stream processing accelerators. 22nd Reconfigurable Architectures Workshop (RAW 2015). USA : IEEE Computer Society, 2015. pp. 81-89
    @inproceedings{abb9ccab8ae94877a776d4b66d8c5de3,
    title = "Real-time multiprocessor architecture for sharing stream processing accelerators",
    abstract = "Stream processing accelerators are often applied in MPSoCs for software defined radios. Sharing of these accelerators between different streams could improve their utilization and reduce thereby the hardware cost but is challenging under real-time constraints. In this paper we introduce entry- and exit-gateways that are responsible for multiplexing blocks of data over accelerators under real-time constraints. These gateways check for the availability of sufficient data and space and thereby enable the derivation of a dataflow model of the application. The dataflow model is used to verify the worst-case temporal behavior based on the sizes of the blocks of data used for multiplexing. We demonstrate that required buffer capacities are non-monotone in the block size. Therefore, an ILP is presented to compute minimum block sizes and sufficient buffer capacities. The benefits of sharing accelerators are demonstrated using a multi-core system that is implemented on a Virtex 6 FPGA. A stereo audio stream from a PAL video signal is demodulated in this system in real-time where two accelerators are shared within and between two streams. In this system sharing reduces the number of accelerators by 75{\%} and reduced the number of logic cells with 63{\%}.",
    keywords = "Real-time applications, multiprocessing systems, stream processing accelerators, METIS-312627, IR-96052, Data flow, accelerator sharing, EWI-26067",
    author = "B.H.J. Dekens and Bekooij, {Marco Jan Gerrit} and Smit, {Gerardus Johannes Maria}",
    note = "10.1109/IPDPSW.2015.147",
    year = "2015",
    month = "5",
    day = "25",
    doi = "10.1109/IPDPSW.2015.147",
    language = "Undefined",
    isbn = "978-1-4673-7684-6",
    publisher = "IEEE Computer Society",
    pages = "81--89",
    booktitle = "22nd Reconfigurable Architectures Workshop (RAW 2015)",
    address = "United States",

    }

    Dekens, BHJ, Bekooij, MJG & Smit, GJM 2015, Real-time multiprocessor architecture for sharing stream processing accelerators. in 22nd Reconfigurable Architectures Workshop (RAW 2015). IEEE Computer Society, USA, pp. 81-89. https://doi.org/10.1109/IPDPSW.2015.147

    Real-time multiprocessor architecture for sharing stream processing accelerators. / Dekens, B.H.J.; Bekooij, Marco Jan Gerrit; Smit, Gerardus Johannes Maria.

    22nd Reconfigurable Architectures Workshop (RAW 2015). USA : IEEE Computer Society, 2015. p. 81-89.

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

    TY - GEN

    T1 - Real-time multiprocessor architecture for sharing stream processing accelerators

    AU - Dekens, B.H.J.

    AU - Bekooij, Marco Jan Gerrit

    AU - Smit, Gerardus Johannes Maria

    N1 - 10.1109/IPDPSW.2015.147

    PY - 2015/5/25

    Y1 - 2015/5/25

    N2 - Stream processing accelerators are often applied in MPSoCs for software defined radios. Sharing of these accelerators between different streams could improve their utilization and reduce thereby the hardware cost but is challenging under real-time constraints. In this paper we introduce entry- and exit-gateways that are responsible for multiplexing blocks of data over accelerators under real-time constraints. These gateways check for the availability of sufficient data and space and thereby enable the derivation of a dataflow model of the application. The dataflow model is used to verify the worst-case temporal behavior based on the sizes of the blocks of data used for multiplexing. We demonstrate that required buffer capacities are non-monotone in the block size. Therefore, an ILP is presented to compute minimum block sizes and sufficient buffer capacities. The benefits of sharing accelerators are demonstrated using a multi-core system that is implemented on a Virtex 6 FPGA. A stereo audio stream from a PAL video signal is demodulated in this system in real-time where two accelerators are shared within and between two streams. In this system sharing reduces the number of accelerators by 75% and reduced the number of logic cells with 63%.

    AB - Stream processing accelerators are often applied in MPSoCs for software defined radios. Sharing of these accelerators between different streams could improve their utilization and reduce thereby the hardware cost but is challenging under real-time constraints. In this paper we introduce entry- and exit-gateways that are responsible for multiplexing blocks of data over accelerators under real-time constraints. These gateways check for the availability of sufficient data and space and thereby enable the derivation of a dataflow model of the application. The dataflow model is used to verify the worst-case temporal behavior based on the sizes of the blocks of data used for multiplexing. We demonstrate that required buffer capacities are non-monotone in the block size. Therefore, an ILP is presented to compute minimum block sizes and sufficient buffer capacities. The benefits of sharing accelerators are demonstrated using a multi-core system that is implemented on a Virtex 6 FPGA. A stereo audio stream from a PAL video signal is demodulated in this system in real-time where two accelerators are shared within and between two streams. In this system sharing reduces the number of accelerators by 75% and reduced the number of logic cells with 63%.

    KW - Real-time applications

    KW - multiprocessing systems

    KW - stream processing accelerators

    KW - METIS-312627

    KW - IR-96052

    KW - Data flow

    KW - accelerator sharing

    KW - EWI-26067

    U2 - 10.1109/IPDPSW.2015.147

    DO - 10.1109/IPDPSW.2015.147

    M3 - Conference contribution

    SN - 978-1-4673-7684-6

    SP - 81

    EP - 89

    BT - 22nd Reconfigurable Architectures Workshop (RAW 2015)

    PB - IEEE Computer Society

    CY - USA

    ER -

    Dekens BHJ, Bekooij MJG, Smit GJM. Real-time multiprocessor architecture for sharing stream processing accelerators. In 22nd Reconfigurable Architectures Workshop (RAW 2015). USA: IEEE Computer Society. 2015. p. 81-89 https://doi.org/10.1109/IPDPSW.2015.147