A multi-grain parallelizing compilation scheme for OSCAR (Optimally scheduled advanced multiprocessor)

Hironori Kasahara, H. Honda, A. Mogi, A. Ogura, K. Fujiwara, S. Narita

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

    11 Citations (Scopus)

    Abstract

    This paper proposes a multi-grain parallelizing compilation scheme for Fortran programs. The scheme hierarchically exploits parallelism among coarse grain tasks, such as, loops, subroutines or basic blocks, among medium grain tasks like loop iterations and among near fine grain tasks like statements. Parallelism among the coarse grain tasks called the macrotasks is exploited by carefully analyzing control dependences and data dependences. The macrotasks are dynamically assigned to processor clusters to cope with run-time uncertainties, such as, conditional branches among the macrotasks and variation of execution time of each macrotask. The parallel processing of macrotasks is called the macro-dataflow computation. A macrotask composed of a Do-all loop, which is assigned onto a processor cluster, is processed in the medium grain in parallel by processors inside the processor cluster. A macrotask composed of a sequential loop or a basic block is processed on a processor cluster in the near fine grain by using static scheduling. A macrotask composed of subroutine or a large sequential loop is processed by hierarchically applying macro-dataflow computation inside a processor cluster. Performance of the proposed scheme is evaluated on a multiprocessor system named OSCAR. The evaluation shows that the multi-grain parallel processing effectively exploits parallelism from Fortran programs.

    Original languageEnglish
    Title of host publicationLanguages and Compilers for Parallel Computing - 4th International Workshop, Proceedings
    PublisherSpringer-Verlag
    Pages283-297
    Number of pages15
    ISBN (Print)9783540554226
    Publication statusPublished - 1992 Jan 1
    Event4th Workshop on Languages and Compilers for Parallel Computing, 1991 - Santa Clara, United States
    Duration: 1991 Aug 71991 Aug 9

    Publication series

    NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume589 LNCS
    ISSN (Print)0302-9743
    ISSN (Electronic)1611-3349

    Other

    Other4th Workshop on Languages and Compilers for Parallel Computing, 1991
    CountryUnited States
    CitySanta Clara
    Period91/8/791/8/9

    Fingerprint

    Compilation
    Multiprocessor
    Parallelism
    Subroutines
    Parallel Processing
    Data Flow
    Macros
    Data Dependence
    Multiprocessor Systems
    Processing
    Execution Time
    Branch
    Scheduling
    Iteration
    Uncertainty
    Evaluation

    ASJC Scopus subject areas

    • Theoretical Computer Science
    • Computer Science(all)

    Cite this

    Kasahara, H., Honda, H., Mogi, A., Ogura, A., Fujiwara, K., & Narita, S. (1992). A multi-grain parallelizing compilation scheme for OSCAR (Optimally scheduled advanced multiprocessor). In Languages and Compilers for Parallel Computing - 4th International Workshop, Proceedings (pp. 283-297). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 589 LNCS). Springer-Verlag.

    A multi-grain parallelizing compilation scheme for OSCAR (Optimally scheduled advanced multiprocessor). / Kasahara, Hironori; Honda, H.; Mogi, A.; Ogura, A.; Fujiwara, K.; Narita, S.

    Languages and Compilers for Parallel Computing - 4th International Workshop, Proceedings. Springer-Verlag, 1992. p. 283-297 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 589 LNCS).

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

    Kasahara, H, Honda, H, Mogi, A, Ogura, A, Fujiwara, K & Narita, S 1992, A multi-grain parallelizing compilation scheme for OSCAR (Optimally scheduled advanced multiprocessor). in Languages and Compilers for Parallel Computing - 4th International Workshop, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 589 LNCS, Springer-Verlag, pp. 283-297, 4th Workshop on Languages and Compilers for Parallel Computing, 1991, Santa Clara, United States, 91/8/7.
    Kasahara H, Honda H, Mogi A, Ogura A, Fujiwara K, Narita S. A multi-grain parallelizing compilation scheme for OSCAR (Optimally scheduled advanced multiprocessor). In Languages and Compilers for Parallel Computing - 4th International Workshop, Proceedings. Springer-Verlag. 1992. p. 283-297. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
    Kasahara, Hironori ; Honda, H. ; Mogi, A. ; Ogura, A. ; Fujiwara, K. ; Narita, S. / A multi-grain parallelizing compilation scheme for OSCAR (Optimally scheduled advanced multiprocessor). Languages and Compilers for Parallel Computing - 4th International Workshop, Proceedings. Springer-Verlag, 1992. pp. 283-297 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
    @inproceedings{d10656d46bdb4e5c8220d16e542ad1a4,
    title = "A multi-grain parallelizing compilation scheme for OSCAR (Optimally scheduled advanced multiprocessor)",
    abstract = "This paper proposes a multi-grain parallelizing compilation scheme for Fortran programs. The scheme hierarchically exploits parallelism among coarse grain tasks, such as, loops, subroutines or basic blocks, among medium grain tasks like loop iterations and among near fine grain tasks like statements. Parallelism among the coarse grain tasks called the macrotasks is exploited by carefully analyzing control dependences and data dependences. The macrotasks are dynamically assigned to processor clusters to cope with run-time uncertainties, such as, conditional branches among the macrotasks and variation of execution time of each macrotask. The parallel processing of macrotasks is called the macro-dataflow computation. A macrotask composed of a Do-all loop, which is assigned onto a processor cluster, is processed in the medium grain in parallel by processors inside the processor cluster. A macrotask composed of a sequential loop or a basic block is processed on a processor cluster in the near fine grain by using static scheduling. A macrotask composed of subroutine or a large sequential loop is processed by hierarchically applying macro-dataflow computation inside a processor cluster. Performance of the proposed scheme is evaluated on a multiprocessor system named OSCAR. The evaluation shows that the multi-grain parallel processing effectively exploits parallelism from Fortran programs.",
    author = "Hironori Kasahara and H. Honda and A. Mogi and A. Ogura and K. Fujiwara and S. Narita",
    year = "1992",
    month = "1",
    day = "1",
    language = "English",
    isbn = "9783540554226",
    series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
    publisher = "Springer-Verlag",
    pages = "283--297",
    booktitle = "Languages and Compilers for Parallel Computing - 4th International Workshop, Proceedings",

    }

    TY - GEN

    T1 - A multi-grain parallelizing compilation scheme for OSCAR (Optimally scheduled advanced multiprocessor)

    AU - Kasahara, Hironori

    AU - Honda, H.

    AU - Mogi, A.

    AU - Ogura, A.

    AU - Fujiwara, K.

    AU - Narita, S.

    PY - 1992/1/1

    Y1 - 1992/1/1

    N2 - This paper proposes a multi-grain parallelizing compilation scheme for Fortran programs. The scheme hierarchically exploits parallelism among coarse grain tasks, such as, loops, subroutines or basic blocks, among medium grain tasks like loop iterations and among near fine grain tasks like statements. Parallelism among the coarse grain tasks called the macrotasks is exploited by carefully analyzing control dependences and data dependences. The macrotasks are dynamically assigned to processor clusters to cope with run-time uncertainties, such as, conditional branches among the macrotasks and variation of execution time of each macrotask. The parallel processing of macrotasks is called the macro-dataflow computation. A macrotask composed of a Do-all loop, which is assigned onto a processor cluster, is processed in the medium grain in parallel by processors inside the processor cluster. A macrotask composed of a sequential loop or a basic block is processed on a processor cluster in the near fine grain by using static scheduling. A macrotask composed of subroutine or a large sequential loop is processed by hierarchically applying macro-dataflow computation inside a processor cluster. Performance of the proposed scheme is evaluated on a multiprocessor system named OSCAR. The evaluation shows that the multi-grain parallel processing effectively exploits parallelism from Fortran programs.

    AB - This paper proposes a multi-grain parallelizing compilation scheme for Fortran programs. The scheme hierarchically exploits parallelism among coarse grain tasks, such as, loops, subroutines or basic blocks, among medium grain tasks like loop iterations and among near fine grain tasks like statements. Parallelism among the coarse grain tasks called the macrotasks is exploited by carefully analyzing control dependences and data dependences. The macrotasks are dynamically assigned to processor clusters to cope with run-time uncertainties, such as, conditional branches among the macrotasks and variation of execution time of each macrotask. The parallel processing of macrotasks is called the macro-dataflow computation. A macrotask composed of a Do-all loop, which is assigned onto a processor cluster, is processed in the medium grain in parallel by processors inside the processor cluster. A macrotask composed of a sequential loop or a basic block is processed on a processor cluster in the near fine grain by using static scheduling. A macrotask composed of subroutine or a large sequential loop is processed by hierarchically applying macro-dataflow computation inside a processor cluster. Performance of the proposed scheme is evaluated on a multiprocessor system named OSCAR. The evaluation shows that the multi-grain parallel processing effectively exploits parallelism from Fortran programs.

    UR - http://www.scopus.com/inward/record.url?scp=85011367933&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=85011367933&partnerID=8YFLogxK

    M3 - Conference contribution

    AN - SCOPUS:85011367933

    SN - 9783540554226

    T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

    SP - 283

    EP - 297

    BT - Languages and Compilers for Parallel Computing - 4th International Workshop, Proceedings

    PB - Springer-Verlag

    ER -