Fast implementation of e for multioperation concurrent read concurrent write MP-SOCs

    Research output: Contribution to journalArticleScientificpeer-review

    1 Citation (Scopus)

    Abstract

    The e-language is a simple parallel extension of c for a class of emulated shared memory multiprocessor systems on chip (W-SOC). It provides an easy-to-use access to fine-grained parallel programming and exploitation of architectural techniques for that class. Unfortunately, the initial implementation of e introduced quite high parallel construct execution time overheads and was strongly tied to exclusive read exclusive write (EREW) memory access model. In this paper we describe faster implementation of e supporting the latest architectural advances in MP-SOCs realizing the multioperation concurrent read concurrent write (MCRCW) model and fast barrier synchronizations. According to our evaluation the new fast mode in together with the stronger computational model provides a significant boost in parallel construct performance for simple e-programs.
    Original languageEnglish
    Pages (from-to)103-110
    JournalWSEAS Transactions on Computers
    Volume6
    Issue number1
    Publication statusPublished - 2007
    MoE publication typeA1 Journal article-refereed

    Fingerprint

    Data storage equipment
    Parallel programming
    Synchronization
    System-on-chip

    Keywords

    • parallel computing
    • parallel languages
    • optimization
    • PRAM model
    • CRCW
    • active memory

    Cite this

    @article{f4faf85cbbd04bdc9f306fa9579788bb,
    title = "Fast implementation of e for multioperation concurrent read concurrent write MP-SOCs",
    abstract = "The e-language is a simple parallel extension of c for a class of emulated shared memory multiprocessor systems on chip (W-SOC). It provides an easy-to-use access to fine-grained parallel programming and exploitation of architectural techniques for that class. Unfortunately, the initial implementation of e introduced quite high parallel construct execution time overheads and was strongly tied to exclusive read exclusive write (EREW) memory access model. In this paper we describe faster implementation of e supporting the latest architectural advances in MP-SOCs realizing the multioperation concurrent read concurrent write (MCRCW) model and fast barrier synchronizations. According to our evaluation the new fast mode in together with the stronger computational model provides a significant boost in parallel construct performance for simple e-programs.",
    keywords = "parallel computing, parallel languages, optimization, PRAM model, CRCW, active memory",
    author = "Martti Forsell",
    note = "CA2: tk703",
    year = "2007",
    language = "English",
    volume = "6",
    pages = "103--110",
    journal = "WSEAS Transactions on Computers",
    issn = "1109-2750",
    publisher = "WSEAS Press",
    number = "1",

    }

    Fast implementation of e for multioperation concurrent read concurrent write MP-SOCs. / Forsell, Martti.

    In: WSEAS Transactions on Computers, Vol. 6, No. 1, 2007, p. 103-110.

    Research output: Contribution to journalArticleScientificpeer-review

    TY - JOUR

    T1 - Fast implementation of e for multioperation concurrent read concurrent write MP-SOCs

    AU - Forsell, Martti

    N1 - CA2: tk703

    PY - 2007

    Y1 - 2007

    N2 - The e-language is a simple parallel extension of c for a class of emulated shared memory multiprocessor systems on chip (W-SOC). It provides an easy-to-use access to fine-grained parallel programming and exploitation of architectural techniques for that class. Unfortunately, the initial implementation of e introduced quite high parallel construct execution time overheads and was strongly tied to exclusive read exclusive write (EREW) memory access model. In this paper we describe faster implementation of e supporting the latest architectural advances in MP-SOCs realizing the multioperation concurrent read concurrent write (MCRCW) model and fast barrier synchronizations. According to our evaluation the new fast mode in together with the stronger computational model provides a significant boost in parallel construct performance for simple e-programs.

    AB - The e-language is a simple parallel extension of c for a class of emulated shared memory multiprocessor systems on chip (W-SOC). It provides an easy-to-use access to fine-grained parallel programming and exploitation of architectural techniques for that class. Unfortunately, the initial implementation of e introduced quite high parallel construct execution time overheads and was strongly tied to exclusive read exclusive write (EREW) memory access model. In this paper we describe faster implementation of e supporting the latest architectural advances in MP-SOCs realizing the multioperation concurrent read concurrent write (MCRCW) model and fast barrier synchronizations. According to our evaluation the new fast mode in together with the stronger computational model provides a significant boost in parallel construct performance for simple e-programs.

    KW - parallel computing

    KW - parallel languages

    KW - optimization

    KW - PRAM model

    KW - CRCW

    KW - active memory

    UR - http://www.worldses.org/journals/computers/old.htm

    M3 - Article

    VL - 6

    SP - 103

    EP - 110

    JO - WSEAS Transactions on Computers

    JF - WSEAS Transactions on Computers

    SN - 1109-2750

    IS - 1

    ER -