id: 05476669
dt: j
an: 05476669
au: Aida, Kento; Iwasaki, Kiyoshi; Okamoto, Masami; Kasahara, Hironori; Narita,
    Seinosuke
ti: Performance evaluation of fortran coarse grain parallel processing on
    shared memory multi-processor systems.
so: Trans. Inf. Process. Soc. Japan 37, No. 3, 418-429 (1996).
py: 1996
pu: Information Processing Society of Japan, Tokyo
la: EN
cc: D.4.1 C.1.2 D.4.8 C.4 D.3.4
ut: performance evaluation; macrodataflow computation scheme; coarse grain
    parallel processing; shared memory multi-processor systems; compiler;
    dynamic scheduling; multi-tasking
ci: 
li: 
ab: Summary: The paper presents an implementation and performance evaluation of
    the macrodataflow computation scheme that is a coarse grain parallel
    processing scheme on shared memory multi-processor systems. In
    macrodataflow computation, the compiler automatically generates coarse
    grain tasks called macrotasks, exploits parallelism among macrotasks
    and generates a dynamic scheduling routine to assign macrotasks to
    processors. A conventional coarse grain parallel processing scheme such
    as multi-tasking has drawbacks such as difficulty in the extraction of
    parallelism among coarse graintasks by users and large dynamic
    scheduling overhead caused by OS calls. However, in macrodataflow
    computation, the compiler exploits parallelism among macrotasks
    automatically and dynamic scheduling overhead is small because the
    compiler optimizes a dynamic scheduling routine for each source
    program. Performance evaluation of the macrodataflow computation scheme
    on an Alliant FX/4 and a Kendall Square Research KSR1 shows that
    macrodataflow computation reduces execution time of programs to
    1/1.91‒1/8.10 of execution time by conventional schemes such as loop
    parallelization and multi-tasking.
rv: